High-resolution imaging of dietary lipids in cells and tissues by NanoSIMS analysis.

Nanoscale secondary ion MS (NanoSIMS) imaging makes it possible to visualize stable isotope-labeled lipids in cells and tissues at 50 nm lateral resolution. Here we report the use of NanoSIMS imaging to visualize lipids in mouse cells and tissues. After administering stable isotope-labeled fatty acids to mice by gavage, NanoSIMS imaging allowed us to visualize neutral lipids in cytosolic lipid droplets in intestinal enterocytes, chylomicrons at the basolateral surface of enterocytes, and lipid droplets in cardiomyocytes and adipocytes. After an injection of stable isotope-enriched triglyceride-rich lipoproteins (TRLs), NanoSIMS imaging documented delivery of lipids to cytosolic lipid droplets in parenchymal cells. Using a combination of backscattered electron (BSE) and NanoSIMS imaging, it was possible to correlate the chemical data provided by NanoSIMS with high-resolution BSE images of cell morphology. This combined imaging approach allowed us to visualize stable isotope-enriched TRLs along the luminal face of heart capillaries and the lipids within heart capillary endothelial cells. We also observed examples of TRLs within the subendothelial spaces of heart capillaries. NanoSIMS imaging provided evidence of defective transport of lipids from the plasma LPs to adipocytes and cardiomyocytes in mice deficient in glycosylphosphatidylinositol-anchored HDL binding protein 1.

Studies with 13 C-and 2 H-labeled lipids and LPs 13 C-uniformly labeled algal fatty acids (Sigma-Aldrich 487937) or 2 H-uniformly labeled stearic acid (Sigma-Aldrich 448249) was dissolved in 200 l ethanol and 1.7 ml sunfl ower oil. These lipids were administered to either wild-type ( Gpihbp1 Ϫ / Ϫ mice. We typically administered 24 mg of 13 C-labeled fatty acids or 18 mg of 2 H-labeled stearic acid for each gavage treatment. In other experiments, we injected mice (generally by a tail vein, rarely intracardiac) with TRLs that were enriched in 13 C-or 2 Hlabeled lipids. To produce stable isotope-enriched TRLs, 13 C algal fatty acids were given to Gpihbp1 Ϫ / Ϫ mice by gavage, typically twice a day over 4 days. The mice were then exsanguinated, yielding ‫ف‬ 1.0 ml of whole blood, which was collected into EDTAcoated tubes. The plasma (0.5 ml) was placed under 2.5 ml of PBS in a 3 ml ultracentrifugation tube and spun at 100,000 rpm for 3 h at 10°C in a TLA 100.3 fi xed angle rotor (Beckman). The TRLs were collected from the top of the tube and placed under 2.5 ml of PBS in a new ultracentrifugation tube and spun again at 100,000 rpm for 1 h. The TRLs were retrieved in a volume of 0.5 ml. Mice were injected with 50-200 l of stable isotopeenriched TRLs containing 300-900 g triglycerides; after 5 or 15 min, the mice were euthanized, and tissue sections were prepared for BSE and NanoSIMS imaging.
In some studies, antibody 11A12 was injected with TRLs.

Preparation of tissue sections
Tissues for NanoSIMS were fi xed with 2.5% glutaraldehyde containing 2 mM MgCl 2 in 100 mM cacodylate buffer (pH 7.4) ( 15,19 ). The samples were then washed three times in distilled water (10 min each). Samples were then treated with 1% osmium tetroxide (OsO 4 ) in 100 mM cacodylate buffer for 1 h, washed in distilled water four times (10 min each), and then incubated with 2% aqueous uranyl acetate overnight at 4°C in the dark. The samples were sequentially dehydrated with increasing concentrations of acetone (20,30,50,70,90, and 100%) for 30 min each, With NanoSIMS imaging, a primary ion beam (Cs + or O Ϫ ) is used to bombard a "surface of interest" (e.g., a section of mouse tissue), releasing secondary ions that can be detected by a magnetic mass analyzer ( Fig. 1 ) ( 10 ). The NanoSIMS instrument detects up to seven distinct masses in parallel. Due to a special lens design and the characteristics of the mass analyzer, these instruments have an unusual combination of high lateral resolution, high mass resolution, and high sensitivity ( 11 ). As with other MS approaches, the NanoSIMS has the ability to detect all isotopes of each element within the sample. This characteristic makes it possible to image and quantify stable isotope-labeled lipids in tissues at high lateral resolution. An advantage of using stable isotope-labeled lipids for imaging is that the stable isotopes have little or no effect on the biochemical properties of the lipids.
Here we describe NanoSIMS imaging for visualizing stable isotope-labeled lipids in cells and tissues ( Fig. 2 ). After feeding mice either 13 C-or 2 H-labeled fatty acids, we harvested tissues and imaged tissue sections with a Nano-SIMS 50 instrument. We also visualized lipids in mice after an intravenous injection of triglyceride-rich LPs (TRLs) enriched in 13 C or 2 H lipids. In many of our studies, we performed backscattered electron (BSE) imaging and NanoSIMS imaging on the same tissue sections. BSE images provide high-resolution morphological information, making it possible to correlate the chemical information provided by NanoSIMS imaging to subcellular features of individual cells. NanoSIMS imaging was effective in documenting defective lipid transport in the setting of glycosylphosphatidylinositol-anchored HDL binding protein 1 (GPIHBP1) defi ciency (where LPL does not reach the capillary lumen and the intravascular processing of TRLs is impaired).

Mouse models
Gpihbp1 knockout mice ( Gpihbp1 Ϫ / Ϫ ) mice have been described previously ( 12 ). All mice were fed a chow diet and housed in a barrier facility with a 12 h light-dark cycle. On a chow diet, Gpihbp1 Ϫ / Ϫ mice have plasma triglyceride levels between 2,500 and 4,000 mg/dl (12)(13)(14)(15). All animal studies were approved by University of California, Los Angeles's Animal Research Committee. showing the focused primary ion beam and the collection and detection of secondary ion signals. A: A Cs + or O Ϫ beam is used to bombard the surface of a sample (e.g., a tissue section), and secondary particles are released from the surface. Among these secondary particles, there are charged ions that can be detected by MS. B: The secondary ions from the surface of the sample pass through a secondary ion column and are analyzed by a Mauttach-Herzog confi guration mass analyzer, which detects secondary ions with high sensitivity and high mass resolution.
TEM imaging ( 22 ). OsO 4 -stained regions in tissue sections generate more BSEs than nonstained regions of the tissue sections; hence, tissue sections stained with OsO 4 are ideal for BSE imaging. The resolution of BSEs is limited by the size of the interaction volume, which can be modulated by the accelerating voltage of the primary electron beam ( 23 ). A low-energy primary electron beam is used to achieve high-resolution BSE imaging. In our studies, resin-embedded, OsO 4 -stained tissue sections were fi rst imaged by optical microscopy. Areas of interest were identifi ed, and the sections were then transferred to an NVision FIB scanning electron microscope for BSE imaging. BSE images were taken with a 2 kV incident beam with a standard aperture (30 µm) and a 5 mm working distance between objective lens and the tissue specimen.

NanoSIMS imaging
SIMS imaging is based on MS of ionized secondary particles emitted from a surface of a sample of interest (e.g., a tissue section) during bombardment with primary ions. Due to collisions between the primary ion beam and the sample, a variety of particles (e.g., electrons, molecules, and atomic or cluster ions) are released from the surface of the sample. Most of these particles are neutral. In most SIMS instruments, only secondary ions can be detected unless special measures are taken to achieve postionization of neutral particles ( 24 ). Recent modifi cations of SIMS instruments have focused on increasing sputtering yield and ionization effi ciency to achieve higher sensitivity, increasing the precision of the mass spectrometer, and improving spatial resolution ( 25,26 ). The NanoSIMS 50 instrument (CAMECA, France) is a high-resolution SIMS instrument designed to improve spatial resolution without compromising the high sensitivity required to detect trace analytes. The NanoSIMS uses the same ion matter interaction mechanisms as other SIMS instruments, but the primary ion beam bombards the sample from a perpendicular angle.
The NanoSIMS instrument uses a coaxial design of ion optics, so that the same lens assembly focuses the primary ion beam and extracts secondary ions emitted from the sample. This design has a very short working distance, allowing a small probe size and improved secondary ion transmission. There are two types of primary ion sources for the NanoSIMS, O Ϫ and Cs + , which allow the analysis of positive and negative secondary ions, respectively. Current followed by three additional treatments with 100% acetone for 20 min each. Samples were then infi ltrated with increasing concentrations of Spurr's resin (25% for 1 h, 50% for 1 h, 75% for 1 h, 100% for 1 h, 100% overnight at room temperature) and then incubated overnight at 70°C in a resin mold. Sections 500 nm thick were cut with a Diatome diamond knife on an ultramicrotome. The sections were placed on a platinum-coated coverslip and left to dry. We used chemical fi xation because it is simple and has been used intensively and successfully for in vivo Nano-SIMS studies in mice ( 20,21 ).
TRL-injected tissues for transmission EM (TEM) imaging were incubated in 2% glutaraldehyde and 0.5% tannic acid in 0.1 M PBS (pH 7.4) for 2 h at room temperature, then washed fi ve times in 0.1 M PBS, and postfi xed in 1% osmium tetroxide in PBS. The samples were then washed four times in sodium acetate buffer (pH 5.5) and block stained in 0.5% uranyl acetate in 0.1 M sodium acetate buffer for 12 h at 4°C. Samples were dehydrated in graded ethanol (50%, 75%, 95%, and three times at 100%) for 10 min each, passed through propylene oxide, and infi ltrated in mixtures of Epon 812 and propylene oxide (1:1 and then 2:1 for 2 h each). Samples were then infi ltrated in pure Epon 812 overnight, embedded in pure Epon 812, and cured for 48 h at 60°C. Sections of 60 nm were cut on an ultramicrotome (RMC MTX) with a diamond knife. The sections were deposited on single-hole grids coated with Formvar and carbon and double stained in aqueous solutions of 8% uranyl acetate for 25 min at 60°C and lead citrate for 3 min at room temperature. Sections 70 nm thick were placed on grids and imaged with a 100CX JEOL electron microscope.

BSE imaging
A scanning electron microscope produces BSEs that can be used to create high-resolution images of a tissue section. BSEs are electrons that are refl ected from the surface of a specimen due to elastic interactions of primary electrons with the nuclei of atoms within the tissue section. The fraction of electrons backscattered by heavy elements (high atomic number) is higher than from light elements (low atomic number). Consequently, the BSE signal is able to reveal contrast based on the average atomic number of atoms in the specimen. OsO 4 is often used to stain biological samples for EM because it stains lipids and creates contrast for Fig. 2. NanoSIMS imaging of mouse tissues. Mice were given stable isotope-enriched fatty acids by gavage or were injected with stable isotope-enriched TRLs. Tissues were harvested, fi xed, dehydrated, resin embedded, and sectioned. A tissue section is transferred to a NanoSIMS 50 instrument, which generates images at up to 50 nm lateral resolution. NanoSIMS imaging provides high-resolution chemical information based on the distribution of stable isotopes within the sample. Shown here are 12 C 14 N Ϫ and 13 C/ 12 C NanoSIMS images of brown adipose tissue (BAT) from a wild-type mouse that had been fed 13 C fatty acids for 4 days. In the 12 C 14 N Ϫ image, an erythrocyte within the lumen of a capillary is white (refl ecting a high 14 N content), whereas the cytosolic lipid droplets of adipocytes are black (refl ecting a low 14 N content). In the 13 C/ 12 C ratio image, the lipid droplets are orange, refl ecting enrichment with 13 C. In some of our studies, we initially performed BSE imaging on tissue sections. The BSE imaging was useful for identifying regions of interest for subsequent NanoSIMS imaging. This approach allowed us to correlate high-resolution morphological information from the BSE image with the chemical information provided by the NanoSIMS image.

RESULTS
After administering 13 C fatty acids to wild-type mice by gavage for 4 days, we performed NanoSIMS imaging on duodenal enterocytes. Fig. 3A , B shows mosaics of 12 C 14 N Ϫ NanoSIMS images, while Fig. 3C, D shows mosaics of 13 C/ 12 C ratio images. The 13 C/ 12 C ratio was high in cytosolic lipid droplets (white arrows, Fig. 3C, E ) in the apical portion of enterocytes and in chylomicrons that had been secreted along the basolateral surface of enterocytes (red arrows, Fig. 3D, F ). In the same mice, NanoSIMS imaging demonstrated 13 C enrichment in the cytosolic lipid droplets of BAT ( Fig. 4A , B ) and heart ( Fig. 4C, D ). In these studies, the 13 C content of cytosolic lipid droplets in BAT was ‫ف‬ 2.2%, approximately twice the natural abundance of 13 C. The 13 C content of lipid droplets in the heart was ‫ف‬ 1.8%, while the 13 C content in heart cytosol (apart from lipid droplets) was 1.16%. Fig. 4E shows NanoSIMS images NanoSIMS instruments can achieve spatial resolutions of 50 nm for a 0.3 pA Cs + primary ion beam and 200 nm pA for a 0.3 pA O Ϫ primary ion beam ( 27 ). An important design feature of the NanoSIMS is the mass analyzer in a Mauttach-Herzog confi guration with a 90° spherical electrostatic section and asymmetric magnet. This mass analyzer makes it possible for the NanoSIMS instrument to achieve both high ion transmission (for sensitivity) and high mass resolution (for accurate identifi cation of the secondary ion), combined with high lateral resolution ( 11 ). These properties make it possible to image stable isotopes in tissue sections at a resolution suffi cient to identify subcellular features of cells.
In this study, a Cs + primary beam was used to bombard the sample surface, and negative secondary ions ( were collected and quantifi ed. These secondary ion signals are affected by ionization probability, which in turn is affected by the work function of the surface. Cs + implantation lowers the surface work function, leading to increased production of negatively charged secondary ions. Secondary ion signals can achieve a steady state after a defi ned dose of primary ion implantation. An optimized dose of 1 × 10 17 ions/cm 2 Cs + was selected to achieve reliable quantifi cation of 13 C/ 12 C and 2 H/ 1 H ratios for imaging of lipids in tissue sections. The natural abundance of 13 C is 1.1%, whereas the natural abundance of 2 H is 0.015%. When biological samples have been metabolically labeled with 13 C-or 2 H-labeled lipids, the NanoSIMS is capable of accurately measuring the amounts of 13 C or 2 H above the baseline levels (due to the natural abundance of these isotopes). Thus, when mice are fed 13 C-or 2 H-labeled fatty acids, it is possible to create a high-resolution image of the distribution of 13 C-and 2 Hlabeled lipids in cells and tissues. In a typical experiment, images are generated based on the local 13 C/ 12 C ratio or the 2 H/ 1 H ratio in the specimen. The 12 C 14 N Ϫ signal is collected to assist in defi ning cell morphology. In our studies, 13 C/ 12 C ratios were calculated on ‫ف‬ 2,000 µm 2 of BAT from wild-type mice and 1,000 µm 2 of BAT from Gpihbp1 Ϫ / Ϫ mice; 2,500 µm 2 of liver from wild-type mice and 2,000 µm 2 of liver from Gpihbp1 Ϫ / Ϫ mice; 4,000 µm 2 of heart from wildtype mice and 3,000 µm 2 of heart from Gpihbp1 Ϫ / Ϫ mice. 2 H/ 1 H ratios on heart were generated from 1,200 µm 2 of wild-type and Gpihbp1 Ϫ / Ϫ heart. NanoSIMS images are typically 256 × 256 pixels or 512 × 512 pixels. To obtain a high-quality NanoSIMS image, it is important to match the primary ion beam size and image pixel size. The D1 aperture was used to control the spot size of the primary ion beam. A large primary aperture (D1 = 2) was used to acquire lowresolution images of tissue sections (40 × 40 µm, 256 × 256 pixels). A smaller primary aperture (D1 = 3 or D1 = 4) was used to achieve higher spatial resolution images of capillaries. The OpenMIMS plugin in ImageJ software (MIMS, Harvard University) was used to process NanoSIMS images. The BSE and NanoSIMS imaging are complementary techniques for imaging biological samples and can be performed on the same tissue section. All of the imaging studies shown here were conducted on tissue sections 500 nm thick, which are ideal for both BSE and NanoSIMS imaging. After obtaining BSE images, the tissue sections were then coated with 5 nm platinum and transferred to the NanoSIMS instrument for analysis. The regions of interest from the BSE image can be localized with an optical microscope within the NanoSIMS instrument. Combined BSE and NanoSIMS imaging makes it possible to correlate the chemical information from the NanoSIMS analysis with the ultrastructural morphology provided by BSE imaging. C ratios near the edges of the droplets. This likely can be explained by the techniques that were used. The primary ion beam size is set at ‫ف‬ 1.5 times the pixel size, so it is possible that pixels on the edge of lipid droplets include signals from lipid droplets and non-lipid droplet portions of the cell. Also, a median fi lter was used to reduce the noises in images; that might also contribute to the lower enrichment ratios at the edges of lipid droplets. lipid droplets (0.32%) was much higher than its natural abundance (0.015%) ( Fig. 6B ). The 2 H Ϫ signal was weaker in cardiomyocytes from Gpihbp1 Ϫ / Ϫ mice ( Fig. 6C, D ). As expected, we observed a high 2 H/ 1 H ratio in the heart capillary lumen of the Gpihbp1 Ϫ / Ϫ mouse ( Fig. 6D ), refl ecting defective processing of 2 H TRLs.
Weinstein et al. ( 28 ) proposed that reduced uptake of TRL lipids in peripheral tissues (e.g., adipose tissue and heart) of Gpihbp1 Ϫ / Ϫ mice is accompanied by increased lipid uptake in the liver. To assess the validity of that concept, we examined the liver of Gpihbp1 +/+ and Gpihbp1 Ϫ / Ϫ mice that had been given 13 C-labeled fatty acids by gavage for 4 days ( Fig. 7 ). In Gpihbp1 +/+ hepatocytes, the lipid droplets were small and 13 C enrichment (1.25%) was minimal. In Gpihbp1 Ϫ / Ϫ mice, lipid droplets were larger, and the 13 C content of lipid droplets was much greater (1.68%). The 13 C content in other regions of the cytoplasm (apart from obvious lipid droplets) was 1.34% in Gpihbp1 Ϫ / Ϫ hepatocytes versus 1.24% in Gpihbp1 +/+ hepatocytes.
To defi ne the localization of stable isotope-enriched lipids, we performed both NanoSIMS and BSE imaging on tissue sections from wild-type mice that had been given an intravenous injection of 13 C-or 2 H-enriched TRLs and then euthanized 5 or 15 min later. By BSE imaging, we observed large spherical particles on the luminal surface of capillaries ( Fig. 8A , C ). Subsequent NanoSIMS imaging on the same section confi rmed that these particles (identifi ed by red arrows) were TRLs enriched in 13 C ( Fig. 8B ) or 2 H ( Fig. 8D ). In Fig. 8D , we detected 2 H enrichment in a cytosolic lipid droplet of a cardiomyocyte (white arrow), indicating that TRL lipids can move from the plasma compartment to parenchymal cells within 15 min.
Next, we investigated whether NanoSIMS imaging was able to detect GPIHBP1 (an endothelial cell protein that binds and transports LPL) on the luminal surface of capillary after an intravenous injection of 13 C-enriched TRLs; the 13 C content of cytosolic lipid droplets of cardiomyocytes in these studies was 1.37%.
To determine whether NanoSIMS imaging can detect physiologically important perturbations in plasma triglyceride metabolism, we fed Gpihbp1 +/+ and Gpihbp1 Ϫ / Ϫ mice 13 C fatty acids by gavage twice daily for 4 days. BAT and heart were harvested, and sections were prepared for NanoSIMS imaging. The analysis revealed more 13 C enrichment in BAT lipid droplets of wild-type mice than in BAT lipid droplets from Gpihbp1 Ϫ / Ϫ mice ( Fig. 5A , B ). The 13 C/ 12 C ratio in lipid droplets of wild-type BAT was 2.2% versus 1.45% in Gpihbp1 Ϫ / Ϫ BAT ( Fig. 5C ). The situation was similar in heart ( Fig. 5D, E ). The 13 C/ 12 C ratio in cardiomyocyte lipid droplets was 1.8% in wild-type mice versus 1.46% in Gpihbp1 Ϫ / Ϫ mice ( Fig. 5F ). The 13 C/ 12 C ratios do not fully refl ect the extent of the metabolic defect in Gpihbp1 Ϫ / Ϫ heart because the numbers of lipid droplets in cardiomyocytes are markedly reduced in Gpihbp1 Ϫ / Ϫ mice (supplementary Fig. I). Of note, the 13 C/ 12 C ratio of the capillary lumen of Gpihbp1 Ϫ / Ϫ mice was much higher than in wild-type mice, refl ecting a striking accumulation of 13 C TRLs in the plasma of Gpihbp1 Ϫ / Ϫ mice (due to defective LPL-mediated processing of the TRLs) ( Fig. 5C, F ). 13 C is abundant in nature (1.1%), resulting in a 13 C Ϫ signal in tissue sections of wild-type mice that have not been given any 13 C-labeled lipids (supplementary Fig. II). Deuterium has a far lower natural abundance (0.015%), resulting in a much lower background of 2 H Ϫ secondary ions (supplementary Fig. III). For this reason, we fed Gpihbp1 +/+ and Gpihbp1 Ϫ / Ϫ mice a commercially available deuterated fatty acid ([ 2 H]stearic acid) twice per day by gavage for 4 days and then performed NanoSIMS imaging on heart tissue. We observed a strong 2 H Ϫ signal in wildtype heart ( Fig. 6A ); the 2 H enrichment in cardiomyocyte Ϫ and 13 C/ 12 C ratio images of BAT from a wild-type mouse that had been given 13 C fatty acids by gavage twice daily for 4 days. C, D: Ϫ and 13 C/ 12 C ratio images of heart from a wild-type mouse that was injected with 13 C TRLs. In E, arrows indicate subtle regions of 13 C enrichment corresponding to lipid droplets. Scale bars: 10 µm (A); 2 µm (B); 15 µm (C); 4 µm (D); 3 µm (E).
inside a capillary endothelial cell vesicle and in TRLs within the capillary lumen.
We took the opportunity to examine heart capillaries of Cav1 -defi cient mice ( Fig. 11E-J ). Cav1 -defi cient mice have reduced numbers of caveolae on capillary endothelial cells ( 19 ), but from the onset, we assumed that our chances of fi nding obvious differences in the handling of TRLs was endothelial cells. We injected wild-type and Gpihbp1 Ϫ / Ϫ mice with a 15 N-labeled rat monoclonal antibody (11A12) against GPIHBP1. After 15 min, the mice were perfused with PBS, and sections of heart were prepared for imaging. After identifying capillaries by BSE imaging, Nano-SIMS analysis was performed on the same sections. We observed localized regions of 15 N enrichment along the luminal face of heart capillaries in wild-type mice, refl ecting binding of the 15 N monoclonal antibody to GPIHBP1 ( Fig. 9A ). As expected, no 15 N enrichment was detected in heart capillaries of Gpihbp1 Ϫ / Ϫ mice (refl ecting the absence of GPIHBP1 on endothelial cells and the absence of 15 N-antibody binding) ( Fig. 9B ). When we image heart capillaries of wild-type mice by TEM, we often observe dark-staining TRLs along the luminal face of endothelial cells ( Fig. 10A , B ). By BSE, we were able to obtain high-resolution images of darkly stained TRLs in capillary endothelial cells ( Fig. 10C, D ). By TEM, we have observed subendothelial TRLs in the capillaries of a wild-type mouse heart 15 min after injection of TRLs ( Fig. 10E, F ). To further explore these fi ndings, we performed correlative BSE and NanoSIMS imaging on heart capillaries. In Fig. 11A , B , we show BSE and 13 C/ 12 C Nano-SIMS images of a heart capillary from a wild-type mouse that had been given 13 C fatty acids by gavage 4 h earlier. A discrete area of 13 C enrichment was observed within a heart capillary endothelial cell ( Fig. 11B ). In Fig. 11C, D , we show BSE and 13 C/ 12 C NanoSIMS images of a capillary from a wild-type mouse 15 min after an intravenous injection of 13 C-labeled TRLs. 13 C enrichment was observed   Ϫ . However, one can easily compensate for the inability low. In our earlier studies ( 19 ), we showed that Cav1 deficiency has little or no impact on GPIHBP1 traffi cking across endothelial cells or on pre-or postheparin plasma LPL levels ( 19 ). In Fig. 11E-J , we show BSE images and 13 C/ 12 C NanoSIMS images of heart capillaries of Cav1 -defi cient mice 15 min after an intravenous injection of 13 C TRLs. Once again, 13 C lipids were observed inside capillary endothelial cells (corresponding to lipid-staining material in the BSE images).
In Fig. 11K, L , we show BSE and 13 C/ 12 C NanoSIMS images of a heart capillary from a wild-type mouse 15 min after an intravenous injection of 13 C TRLs. In these images, we observed a 13 C-labeled TRL in the subendothelial space and 13 C-enriched TRLs that had marginated along the luminal face of the capillary endothelial cells.

DISCUSSION
In the current study, we describe NanoSIMS imaging of stable isotope-enriched lipids in cells and tissues. We fed mice either 13 C-or 2 H-labeled fatty acids or injected mice with TRLs containing 13 C or 2 H lipids. We were able to document the uptake and secretion of stable isotope-labeled lipids in the intestine, the movement of lipids across endothelial cells, and the uptake and the subcellular localization of lipids within adipocytes and cardiomyocytes. NanoSIMS imaging worked well with both 13 C-and 2 H-labeled fatty acids, but the amount of stable isotope enrichment, relative to natural abundance, was greater with 2 H-labeled lipids. For that reason, studies with 2 H fatty acids are probably better suited for visualizing small amounts of stable isotope enrichment in tissues (e.g., discerning low levels of stable isotope enrichment at early time  of a metabolically labeled monoclonal antibody in tissue sections. Our approach is simpler than making antibodies labeled with fl uorinated gold beads, and one does not need to worry about an adverse effect of bead conjugation on antibody binding. We anticipate that there will be many opportunities to use NanoSIMS imaging to visualize antibody binding in vivo ( 32 ). In the realm of cancer medicine, NanoSIMS imaging could prove to be useful for examining the targeting of diagnostic or therapeutic antibodies to tumor cells.
The ability of NanoSIMS imaging to detect perturbations in lipid metabolism was clearly demonstrated by our studies in Gpihbp1 +/+ and Gpihbp1 Ϫ / Ϫ mice. Normally, myocytes and adipocytes secrete LPL into the interstitial spaces; this LPL is then "swept up" by GPIHBP1 on endothelial cells and shuttled to the capillary lumen ( 19,30,33 ). In the setting of GPIHBP1 defi ciency, the LPL in adipose tissue and striated muscle is mislocalized within the interstitial spaces, preventing LPL from interacting with TRLs and leading to both hypertriglyceridemia and impaired delivery of lipid nutrients to parenchymal cells ( 12,19,28,30,33 ). Our NanoSIMS studies provide visual documentation of the perturbed triglyceride metabolism in Gpihbp1 Ϫ / Ϫ mice, revealing an accumulation of stable Ϫ secondary ions in a single scan. NanoSIMS imaging was useful for visualizing the binding of a 15 N-labeled GPIHBP1-specifi c antibody (11A12) to the luminal face of capillaries. In earlier experiments, we showed that 124 I-11A12, when injected intravenously into wild-type mice, binds to GPIHBP1 on endothelial cells and disappears from the bloodstream within 1 min ( 29 ). Similarly, a fl uorescently labeled antibody 11A12, when injected intravenously, quickly binds to the luminal surface of capillaries, as judged by confocal microscopy on tissue sections 10 m thick ( 19,30 ). By confocal microscopy, the binding of the fl uorescently labeled antibody was patchy but detectable along the entire circumference of the capillary lumen. In the NanoSIMS studies reported here, the 15 N-labeled monoclonal antibody was observed in discrete patches along capillaries but was not detected along the entire circumference of the capillary. This is similar to the pattern of GPIHBP1 distribution in heart capillaries by immunogold EM ( 19 ). We suspect that fi nding only small patches of 15 N enrichment in the NanoSIMS images is due to the fact that we collected secondary ions from an extremely thin layer of the tissue section ( ‫ف‬ 10 nm). In earlier studies, Wilson et al. ( 31 ) used NanoSIMS imaging to detect a monoclonal antibody conjugated to fl uorinated gold beads; antibody binding was documented by Nano-SIMS by collecting 19 F Ϫ secondary ions. As far as we are aware, our experiments are the fi rst to show that Nano-SIMS imaging is suffi ciently sensitive to detect the binding level of 13 C enrichment in Gpihbp1 Ϫ / Ϫ hepatocytes strongly support that view. In our studies, we often performed BSE and NanoSIMS imaging on the same tissue section. This approach made it possible to localize regions of interest to examine by NanoSIMS imaging and made it possible to correlate the chemical information from the NanoSIMS with the highresolution morphological information from the BSE imaging. This correlative approach was quite useful. First, we were able to show that most of the 13 C or 2 H lipid in parenchymal cells was concentrated in neutral lipid droplets; 13 C or 2 H enrichment in other areas of the cell was low. In the 13 C experiments, the 13 C content in the cytosol (apart from lipid droplets) was only slightly greater than the natural abundance of the isotope. Second, correlative BSE and NanoSIMS imaging studies allowed us to identify 13 Cor 2 H-enriched TRLs that had marginated along the luminal surface of capillary endothelial cells. Third, the correlative methodologies allowed us to defi nitively identify discrete areas of 13 C enrichment within capillary endothelial cells.
isotope-enriched TRLs in the capillary lumen and reduced delivery of TRL lipids to parenchymal cells.
NanoSIMS imaging of the liver was intriguing, revealing a higher 13 C/ 12 C ratio in lipid droplets and cytoplasm in Gpihbp1 Ϫ / Ϫ mice than in Gpihbp1 +/+ mice. Normally, the adult mouse liver expresses only small amounts of LPL ( 12,14,34 ). However, the fact that capillaries in the liver are fenestrated means that LPL-mediated processing of TRLs does not require transport of LPL by GPIHBP1 ( 14 ). When the processing of TRLs in adipose tissue and striated muscle is defective, as in GPIHBP1 defi ciency, the liver plays a proportionately greater role in TRL processing ( 28 ). In an earlier study of Gpihbp1 Ϫ / Ϫ mice, Weinstein et al. ( 28 ) found reduced levels of essential fatty acids in the adipose tissue of Gpihbp1 Ϫ / Ϫ mice but greater than normal levels of essential fatty acids in the liver. They proposed that the defective TRL processing in peripheral tissues of Gpihbp1 Ϫ / Ϫ mice is accompanied by a reciprocal increase in TRL processing and lipid uptake in the liver ( 28 ). Our NanoSIMS images showing a greater than normal C NanoSIMS images of a heart capillary from a wild-type mouse that had been given 13 C fatty acids by gavage 4 h earlier. In the NanoSIMS image, a discrete area of 13 C enrichment was observed inside an endothelial cell (arrow). C, D: BSE and 13 C/ 12 C NanoSIMS images of a heart capillary of a wild-type mouse 15 min after an intravenous injection of 13 C TRLs. In the NanoSIMS image, a discrete area of 13 C enrichment was observed inside an endothelial cell vesicle (arrow); 13 C enrichment was also observed in TRLs in the capillary lumen. For the image shown in D, 256 × 256 pixel NanoSIMS images were recorded initially. A small region of interest was identifi ed (guided by the BSE image), and that region was enlarged. E-J: BSE images and 13 C/ 12 C NanoSIMS images of heart capillaries from a Cav1 -defi cient mouse 15 min after an injection of 13 C TRLs. 13 C-enriched lipids (arrows) inside capillary endothelial cells corresponded to dark-staining material on the BSE. In J, other areas of 13 C enrichment corresponded to cytosolic lipid droplets and a TRL that had marginated along the luminal surface of a capillary. K, L: BSE and 13 C/ 12 C NanoSIMS images of a heart capillary from a wildtype mouse 15 min after an intravenous injection of 13 C TRLs. A 13 C TRL was observed in the subendothelial space (red arrow). 13 C enrichment was also detected in TRLs that had marginated along the luminal face of capillary endothelial cells (white arrows). Scale bars: 1 µm (A-H), 1.5 µm (I-L). be intriguing to visualize triglyceride transport across endothelial cells in vertebrate species that lack GPIHBP1 expression, for example birds and fi sh ( 33 ). Finally, it would be interesting to examine fatty acid and glucose delivery to tumor cells and determine whether different cells within the tumor are metabolically homogenous.
In most cases, these areas of enrichment were in "lipid particles" resembling TRLs. Finally, the correlative BSE/ NanoSIMS approach made it possible to identify a stable isotope-enriched TRL within the subendothelial spaces. Subendothelial TRLs were also identifi ed by TEM.
Our studies provide visual evidence of lipid movement from the plasma to endothelial cells and parenchymal cells. However, we are not convinced that we are "seeing" the entirety of the lipid transport process. By TEM, BSE, and NanoSIMS, we have encountered hundreds of TRLs marginated along the luminal surface of capillaries. However, in contrast to the abundance of marginated TRL in capillaries, we found discrete patches of lipids inside endothelial cells rather infrequently. In those cases, the areas of stable isotope enrichment often corresponded to darkly staining material resembling TRLs. In our NanoSIMS imaging studies, we suspect that we have visualized the uptake and transport of intact TRLs by endothelial cells. We know that TRLs bind to the GPIHBP1-LPL complex on capillary endothelial cells (often near caveolar-like invaginations) ( 15,19 ). We also know that GPIHBP1 moves back and forth across endothelial cells ( 19,29 ). We suspect that most examples of stable isotope enrichment inside capillary endothelial cells are due to TRLs "catching a ride" with GPIHBP1. We are not yet convinced that we are adequately visualizing the movement of the fatty acid products of TRL processing. One possibility is that fatty acids move across endothelial cells in vesicles but that the vesicular traffi cking is so rapid that the vesicles are detected only rarely in tissue sections. Another possibility is that the fatty acids diffuse so quickly and so broadly that the stable isotope signal instantly disappears within an "ocean" of naturally occurring 13 C and 2 H atoms within cells and tissues. To address this issue, it would be desirable to perform NanoSIMS imaging in mice after injecting 14 C TRLs. 14 C is essentially absent in nature, meaning that the background in NanoSIMS imaging studies would be nonexistent. With 14 C experiments, one should be able to visualize the transport of fatty acids across endothelial cells and into parenchymal cells, even if fatty acids diffuse broadly and quickly. More than three decades ago, Scow et al. ( 35 ) suggested that the fatty acid products of lipolysis diffuse along the plasma membrane of endothelial cells. If that is truly the case, it ought to be possible to visualize the diffusion of fatty acids along the plasma membrane by 14 C NanoSIMS imaging. Unfortunately, the 14 C fatty acids required for these sorts of studies would be very expensive, and even if cost were not a factor, one would still need to identify a NanoSIMS facility that permits studies with radioisotopes.
We encountered rare examples of TRLs in the subendothelial spaces of heart capillaries. We suspect that the subendothelial TRLs had "hitched a ride" across endothelial cells with GPIHBP1. In previous studies, Bartelt et al. ( 36 ) reported that triglyceride-rich liposome particles containing iron nanobeads move across capillaries in BAT.
We imagine additional opportunities for NanoSIMS imaging in lipid research. NanoSIMS imaging of cholesterol transport from native or modifi ed forms of LDLs to cells of the arterial wall would likely be informative. Also, it would