Lipids in biocalcification: contrasts and similarities between intimal and medial vascular calcification and bone by NMR.

Pathomechanisms underlying vascular calcification biogenesis are still incompletely understood. Biomineral from human atherosclerotic intimal plaques; human, equine, and bovine medial vascular calcifications; and human and equine bone was released from collagenous organic matrix by sodium hydroxide/sodium hypochlorite digestion. Solid-state (13)C NMR of intimal plaque mineral shows signals from cholesterol/cholesteryl esters and fatty acids. In contrast, in mineral from pure medial calcifications and bone mineral, fatty acid signals predominate. Refluxing (chloroform/methanol) intimal plaque calcifications removes the cholesterylic but not the fatty acyl signals. The lipid composition of this refluxed mineral now closely resembles that of the medial and bone mineral, which is unchanged by reflux. Thus, intimal and medial vascular calcifications and bone mineral have in common a pool of occluded mineral-entrained fatty acyl-rich lipids. This population of fatty acid may contain methyl-branched fatty acids, possibly representing lipoprotein particle remnants. Cell signaling and mechanistic parallels between physiological (orthotopic) and pathological (ectopic) calcification are also reflected thus in the NMR spectroscopic fingerprints of mineral-associated and mineral-entrained lipids. Additionally the atherosclerotic plaque mineral alone shows a significant independent pool of cholesterylic lipids. Colocalization of mineral and lipid may be coincidental, but it could also reflect an essential mechanistic component of biomineralization.

and 13 C ( 19 ) NMR assignments have been made in cholesterol-fed rabbit atherosclerotic lesions and lipoproteins, and they have been used to investigate lipid compositions and ratios. In model phospholipid-cholesterol binary mixtures, nonesterifi ed cholesterol decreases phospholipid mobility and lipid NMR signal linewidths ( 20 ). Solid-state 13 C and 31 P NMR, respectively, characterize the organic and the mineral components of human carotid artery plaques ( 21 ). NMR quantifi cation of crystalline cholesterol and hydroxyapatite in intact human atherosclerotic plaques correlates with chemical analyses and shows that some cholesterol takes the form of crystalline cholesterol monohydrate ( 22 ).
In this study, we use chemical treatment of pathologically and naturally calcifi ed tissue to remove extracellular matrix organic material, leaving only the mineral deposit itself, thus allowing detailed characterization of its organic contents. We fi nd fatty acyl lipids that resist chemical digestion and subsequent organic solvent extraction in all the biomineral types. However, cholesterol and cholesteryl esters are abundant only in mineral from calcifi ed intimal vascular plaque, and they can be mobilized by organic solvents, suggesting signifi cant differences in the roles of these two lipid populations in the calcifi cation process. Moreover, the form of acyl fatty acids we identify in both vascular tissue and bone raises further questions about the source and role of this lipid population.

MATERIALS AND METHODS
All human material was obtained and used with full institutional ethical approval. Intimal calcifi cations were obtained from carotid endarterectomies; NMR was performed on four samples, three of which were from different individuals, and one of which was prepared from intimal calcifi cations from three other individuals, pooled. Calcifi cations judged to be predominantly medial from six different individuals were obtained from femoral arteries of diabetic patients after amputation. Intimal calcifi ed atherosclerotic plaque was pretreated with collagenase and elastase as described in detail in Duer et al. ( 16 ). Calcifi ed medial vascular material was frozen at Ϫ 20 C as soon as possible after surgery and subjected to sodium hydroxide and sodium hypochlorite digestion (see below) without further treatment. A single sample of human bone was obtained from the femoral head of a patient undergoing hip replacement surgery. Equine tibial cortical bone (from six different horses ranging in age from fetus to 35 years) and calcifi ed pulmonary artery medial plaque (from fi ve different horses and one cow) was obtained from animals which died naturally or were euthanized humanely for reasons unrelated to this study.
Pure lipids were purchased from Sigma (Poole, Dorset, UK) and used without further purifi cation.
Calcifi cations were released from organic matrix and other connective tissue (collagen and other biomacromolecules) by prolonged digestion in concentrated sodium hydroxide and sodium hypochlorite at 4°C. At intervals of a few weeks, samples were fi ltered, thoroughly washed with distilled water, air dried, and interrogated by 13 C SSNMR. If signals from collagenous proteins were observed, particularly the signature signal at ca. 70 ppm from the ␥ -carbon of hydroxyproline, the sample was resuspended in concentrated sodium hydroxide and sodium hypochlorite for another few weeks. This procedure was repeated enriched in acidic phospholipids, such as phosphatidylserine, that could coordinate calcium ions in the initial mineral nucleation stages ( 8,9 ), and they have been found to be loaded with mineralization-supportive proteins and other macromolecules ( 9 ).
Comparative mass spectrometric lipidomics of nonmineralized human atherosclerotic plaques shows abundant unsaturated cholesteryl esters and some phosphatidylcholines and sphingomyelins, with elevated concentrations of lipid species, particularly cholesteryl esters, in plaque relative to plasma ( 10 ). Cholesteryl esters are nearly always associated with atherosclerosis, but what their involvement is in the calcifi cation process is still a matter of considerable debate. Crystallites of cholesterol are found in close proximity to apatitic mineral in atherosclerotic plaques ( 11 ), and in vitro, apatitic crystals are found to grow from cholesterol crystal surfaces ( 12 ), prompting suggestions of a possible templating role. Conversely, hydroxyapatite seed crystals added to phosphatidyl/cholesterol liposomes caused the liposomes to aggregate, suggesting that the lipid role may be much more complex than simple templating. The situation is further complicated by the fact that most cholesterol in the body is carried in low-density lipoprotein (LDL) particles ( 13 ), which also have their own phospholipid membrane. In turn, LDL particles binding to arterial walls is recognized as a key initiator of atherosclerosis ( 14 ), although their possible roles in calcifi cation of atherosclerotic plaques is less clear.
There is considerable evidence that vascular calcifi cation is associated with phenotypic changes of VSMCs toward osteoblast/chondrocyte-like differentiation ( 15 ). Physiological growth-plate calcifi cation of cartilage and bone may involve MVs, although the detail of that involvement is still a subject of considerable research ( 4 ). By analogy, vascular calcifi cation might also be seen as a "growth phase," albeit a pathological one. There are further similarities between physiological calcifi cation of bone and cartilage and pathological vascular calcifi cation (e.g., the respective compositions and crystallinities of the mineral components and the nature of the biomolecules, such as collagen and sugar species ( 16 ), closely associated with the mineral). Detailed atomic-level characterization and comparison of the lipid associated with bone and pathological vascular biomineral deposits could provide further insight into the role of lipids in both normal and pathological mineralization and, thus, suggest new approaches to regulating mineralization processes.
Solid-state NMR is a powerful technique for characterizing the chemical composition and molecular structure of biological materials. It is often effective with little sample manipulation and extraction. This can be particularly useful for composite mineralized tissue because freeing organic material from the mineral phase can require severe and possibly destructive treatments. NMR methods have been extensively applied to vascular deposits and their molecular precursors. Hamilton et al. identifi ed 13 C signals from cholesteryl esters and the choline head groups of phosphatidylcholines in intact LDL and in arterial tissue containing fi brotic lesions ( 17 ). Numerous lipid 1 H shown in Fig. 1B . Probably because of this, there is more variation across the spectra from the medial calcifi cations with respect to the relative proportions of cholesterol to fatty acid (supplementary Fig. II), but in all cases, there is less cholesterol relative to fatty acid than in any of the pure intimal calcifi cations. To interrogate more samples of pure medial calcifi cations uncontaminated by intimal calcification, we obtained samples from nonprimate species (equine and bovine), which are prone to medial rather than intimal vascular calcifi cation (23)(24)(25). A typical 13 C CP-MAS solid-state NMR spectrum of an equine medial calcifi cation is shown in Fig. 1C . There is much less intersample variation in the 13 C NMR spectral profi le among the nonhuman medial calcifi cations (shown in supplementary Fig. III). With the exception of a signal of low and variable intensity from mineral carbonate ion, all the spectra are dominated by fatty acid signals with no detectable signals from cholesterol.

Bone mineral
To interrogate the properties of nonpathological mineralized material, several samples of equine bone and a sample of adult human bone were also digested with sodium hydroxide and sodium hypochlorite to free the mineral from the collagenous organic matrix. 13 C CP-MAS spectra from a human sample and an equine bone mineral sample are shown in Fig. 2A , B , respectively. They are very similar to each other and to fi ve other equine bone samples (supplementary Fig. IV), with the exception of large variations in the proportion of mineral carbonate ion.

Organic solvent extraction
Changes in the spectral characteristics of intimal plaque mineral and bone on chloroform/methanol refl ux are exemplifi ed in Fig. 3 . Refl ux considerably reduces the signals due to cholesterolic compounds in intimal calcifi cations relative to those from fatty acids. In contrast, the content and proportions of lipid in bone hardly change on refl ux. Liquid-state 13 C spectra of the compounds released from the calcifi cations are shown in Fig. 4 .

DISCUSSION
Mineral from vascular calcifi cations closely resembles bone hydroxyapatite by X-ray powder diffraction and 31 P NMR ( 22 ), consistent with the biological mechanistic parallels between normal osteogenesis and pathological vascular mineralization ( 26,27 ). However, our work shows that, after chemical removal of the collagenous organic matrix, intimal atherosclerotic calcifi cations, on the one hand, and medial calcifi cations and bone, on the other, are strikingly different. All three mineral types produce 13 C NMR signals, indicating the presence of partly unsaturated (signal at ca. 130 ppm) long-chain fatty acids on or in the mineral. However, stripping the collagenous matrix from intimal calcifi cations leaves signifi cant amounts of cholesterol and cholesteryl esters, which can be removed by refl ux in chloroform/methanol. Our spectral assignments until continued digestion resulted in no further spectroscopic changes. Yields of calcifi ed material free of organic matrix and connective tissue material varied widely, between a few milligrams and a few hundreds of milligrams, depending on the mass of original available material.
Lipids were extracted from digested mineral by refl uxing calcifi cations from intimal plaque and bone, in 1:1 chloroform:methanol for 30 min. The solvent was decanted, and the refl uxed mineral was washed twice with fresh chloroform/methanol and air dried. The washings were added to the original decanted refl uxate, which was dried under low vacuum on a rotary evaporator and redissolved in deuterochloroform for liquid-state NMR characterization.

Solid-state NMR
All experiments were performed using standard SSNMR methodology on a Bruker 9.4 Tesla Avance-400 wide bore spectrometer, at frequencies of 400. 13 C chemical shifts were referenced to the signal of the methylene carbon of solid ␣ -glycine at 43.1 ppm relative to tetramethylsilane at 0 ppm. Where there was insuffi cient material to fi ll a 4 mm outer diameter rotor unfi lled volume was taken up by PTFE tape. Number of scans acquired depended on available sample and was generally between 10,000 and 100,000 .

Liquid-state NMR
All 13 C experiments were performed in deuterochlorofom (C 2 HCl 3 ) using standard methodology on a Bruker 9.4 Tesla Avance-500 standard bore spectrometer at frequencies of 500.1 MHz ( 1 H) and 125.6 MHz ( 13 C). Standard pulse-acquire with continuous broadband decoupling was used ( /6 excitation pulses, repetition time 2.25 s). Chemical shifts were referenced to the central component of the deuterochloroform triplet at 77.23 ppm relative to tetramethylsilane at 0 ppm. 13 C CP-MAS spectra of mineral from a calcifi ed human intimal plaque (carotid) and from a human medial (femoral) calcifi cation are shown in Fig. 1A , B, respectively. The spectrum of the mineral from the intimal calcifi cation shows signals from cholesterol-related compounds, fatty acids, and carbonate substituted into the hydroxyapatite matrix. In contrast, the spectrum of mineral from the one medial calcifi cation shown only displays signals from fatty acids and no detectable signal from cholesterol. It was possible to digest and interrogate a total of four intimal calcifi cations, one of which was obtained by pooling three separate earlier samples prior to sodium hydroxide and sodium hypochlorite digestion. The 13 C spectra from each were very similar (see supplementary Fig. I). Results for medial calcifi cations were less consistent because of the presence of both medial and intimal calcifi cations in the vessel walls of most patients and because of our inability to separate these layers, with the exception of the sample fatty acids ( 28 ), of which pristanic acid is a typical example found in mammalian tissues. Formed from metabolism of other branched fatty acids, such as phytanic acid in cell perioxisomes, the role of such branched-chain fatty acids in mammalian physiology is not clear, although they are clearly an energy source; pristanic acid can be taken up by mitochondria and metabolized to carbon dioxide and water. Crucially though, branched fatty acids are not common components of phospholipid membranes, so it seems unlikely that this is the source of this lipid component in these calcifi ed tissues.

Vascular calcifi cations
Interestingly, a detailed atomic-force microscopy study ( 29 ) found an abundance of small ( ‫ف‬ 145 nm) lipid particles decorating collagen microfi brils in bone. The nature of these could not be identifi ed for certain, but their size are based on literature, e.g., Kroon et al. ( 19 ), and comparison with pure model compounds (see supplementary Figs. V and VI). Head group and glyceryl signals from phospholipids, e.g., in MVs or lipoprotein particles, such as phosphatidylcholine and phosphatidylserine, expected between ca. 50 and 70 ppm, are not observed. These phospholipids may not be present in detectable quantities, or signals may be broadened by chemical and structural heterogeneity and immobilization. Although refl ux removes cholesterylic signals from intimal calcifi cations, it leaves signals from long-chain fatty acids. In the pure medial calcifi cations and bone mineral, refl ux mobilizes a small proportion of the total lipid, leaving strong fatty acid signals. Thus, NMR shows all three mineral types associated with pools of mineral-entrained fatty acids resistant to alkali, oxidizing agents, and lipid-solubilizing solvents.
After refl ux, the intimal plaque mineral lipid spectra now resemble those of pure medial calcifi cations and bone mineral. This argues for two distinct pools of hydrophobic lipid in or associated with the intimal mineral: one cholesterylic and accessible to organic solvents and possibly a remnant of lipoprotein precursors, and the other less accessible to organic extraction and consisting predominantly of fatty acyl structures.
The 13 C NMR fatty acid signals for straight-chain fatty acids occur at 180 ppm for the COOH acid carbon, with the chain carbons occurring between 22 and 34 ppm, plus the end methyl group at 14-15 ppm, with very little variation in this distribution of signals ( 28 ) except for the unsaturated C = C carbons, which routinely occur close to 130 ppm. However, the spectra of both intimal and medial mineral and bone show strong signals at 38-39 ppm and ca. 183-185 ppm . These arise from 2-methyl-branched  The liquid-state NMR spectra reveal the composition of the lipids released by refl ux. In the refl uxate from intimal plaque mineral ( Fig. 4A ), each of the olefi nic carbons of unesterifi ed and esterifi ed cholesterol gives rise to a pair of well-resolved distinct signals (C5 and C6, and CE5 and CE6, respectively, in Fig. 4 ). Unesterifi ed cholesterol predominates over esterifi ed cholesterol in the organic solventexposed lipid pool, at least after digestion.
led the authors of that study to propose that they were lipoprotein particles. Methyl-branched fatty acids are transported in lipoprotein particles, so this would be consistent with our fi nding here. We can infer that this fatty acid pool must be entrained within mineral structures. These fatty acids may be relics of lipoprotein particles, which give rise to a host of questions about their possible role in matrix calcifi cation.  The much weaker spectrum of the refl uxate from bone mineral shows a relatively much higher proportion of fatty acid signals, such as from long-chain terminal methyl groups (ca. 14 ppm) and from olefi nic carbons (at ca. 130 ppm). There are no detectable signals around 62 and 69 ppm (which would correspond to the glyceryl carbons of intact triglycerides and phosphatidyl glycerides); it is, of course, possible that the sample preparation has leached these components out of the mineral deposits. Liquid-state spectra of model cholesterylic compounds and glycerides are shown for comparison in supplementary Figs. VII and VIII, respectively. We note there are signals consistent with both 2-methyl (37-40 ppm) and n-methyl-branched fatty acids [n ≠ 2 (36-38, 42 ppm)] . These signals arise from carbons with or next to methyl substitutions, and the net intensity of these signals compared with that between 29 and 30 ppm, which represents the remaining fatty acid chain carbons, suggests that such carbons are relatively abundant; in other words, that there is a preponderance of methyl-substituted fatty acids associated with mineral. A spectrum of pristanic acid, a model methyl branched fatty acid, is shown in supplementary Fig. IX. Neutral (zwitterionic) phospholipids predominate over acidic phospholipids in normally calcifying and calcifi ed tissue; the former, but not the latter, are extractable without demineralization ( 30,31 ), and our results are consistent with this. What is equally interesting is that calcifi ed vascular tissue appears to contain a similar distribution of mineral-associated fatty acids, suggesting that these fatty acids play a role in calcifi cation. A possible synergy between cholesterylic compounds and phospholipids in supporting calcifi cation is supported by model liposome-hydroxyapatite systems ( 32 ); it would be interesting to examine how cholesterylic lipids and lipids containing branched-chain fatty acids compare in this regard.
In conclusion, intimal atherosclerotic vascular calcifi cations contain a pool of cholesterylic lipids and a pool of fatty acyl lipids that resist strong alkali and chemical oxidation. The cholesterylic, but not the fatty acyl, population is readily mobilized by organic solvents. In medial vascular calcifi cations, however, alkali and oxidation leave a predominantly fatty acyl lipid pool closely resembling that in bone mineral and the fatty acyl lipids left behind after organic solvent refl ux in intimal calcifi cations. It is likely that this predominantly fatty acyl pool is common to all three calcifi cation types (intimal, medial vascular, and bone). It is characterized by a relative abundance of methylbranched fatty acids, and we tentatively suggest that these may represent the remnants of lipoprotein particles. The cholesterylic lipids observed only in intimal calcifi cations may play an independent role in mineralization, or their presence may be merely a coincidental function of the abundance of such lipids in atherosclerotic lesions. Thus, solid-state NMR shows that participation of fatty acyl lipids in calcifi cation is another factor common to all three families of mineralized tissue, in spite of their different orthotopic or ectopic origins, locations, and pathological signifi cance.