Histochem Cell Biol (2009) 132:105–115

DOI 10.1007/s00418-009-0587-z

123

ORIGINAL PAPER

Chemical analysis of osmium tetroxide staining in adipose tissue using imaging ToF-SIMS

Dalila Belazi · Santiago Solé-Domènech ·

Björn Johansson · Martin Schalling · Peter Sjövall

Accepted: 9 March 2009 / Published online: 25 March 2009

Springer-Verlag 2009

Abstract Osmium tetroxide (OsO4) is a commonly used

stain for unsaturated lipids in electron and optical micros-

copy of cells and tissues. In this work, the localization of

osmium oxide and speciWc lipids was independently moni-

tored in mouse adipose tissue by using time-of-Xight sec-

ondary ion mass spectrometry with Bi cluster primary ions.

Substance-speciWc ion images recorded after OsO4 staining

showed that unsaturated C18 fatty acids were colocalized

with osmium oxide, corroborating the view that osmium

tetroxide binds to unsaturated lipids. In contrast, saturated

fatty acids (C14, C16 and C18) and also unsaturated C16

fatty acids show largely complementary localizations to

osmium oxide. Furthermore, the distributions of saturated

and unsaturated diglycerides are consistent with the speciWc

binding of osmium oxide to unsaturated C18 fatty acids.

The abundance of ions, characteristic of phospholipids and

proteins, is strongly decreased as a result of the osmium

staining, suggesting that a large fraction of these com-

pounds are removed from the tissue during this step, while

ions related to fatty acids, di- and triglycerides remain

strong after osmium staining. Ethanol dehydration after

osmium staining results in more homogeneous distributions

of osmium oxide and unsaturated lipids. This work pro-

vides detailed insight into the speciWc binding of osmium

oxide to diVerent lipids.

Keywords ToF-SIMS · Saturated fatty acids ·

Unsaturated fatty acids · Glycerides · Osmium tetroxide

Introduction

Very little was known about the mechanism of biological

stains when they were introduced for light microscopy in

the seventeenth and eighteenth centuries. During the twen-

tieth century, a rich literature aimed at describing the chem-

istry underlying biological stains was published. For many

stains, however, knowledge of their chemical reactions is

still limited or absent (Zollinger 2003). One obstacle to

such studies is that many methods of chemical analysis

require sample processing (e.g., tissue homogenization)

diVerent from that in histology, thus making a direct com-

parison of chemical and histological results diYcult.

OsO4 is extensively used to stain and Wx tissues for elec-

tron microscopy (Alberts et al. 2007; Moore and Zouridakis

2003; Studer et al. 2008). Osmium is highly electropositive

with an initial oxidation state of +8 and gives rise to strong

electron scattering from electron donor ligands, making this

staining method suitable for electron microscopy (Bozzola

and Russell 1999). The reactivity of cell bio-components

with OsO4 has been explained by the osmium ester forma-

tion theory (Collin and GriYth 1974), according to which

OsO4 reacts with alkenes to form cyclic esters, which sub-

sequently may undergo hydrolysis to give a vicinal diol and

release of a reduced osmium oxide (see Fig. 1). Osmium is

likely to form stable mono-ester and/or di-ester osmiumVII

complexes (Fig. 1a), although di-ester formation is likely to

be the predominant structure in stained tissue (Fig. 1b)

D. Belazi (&) · S. Solé-Domènech · B. Johansson · M. Schalling

Department of Molecular Medicine and Surgery,

Karolinska Institutet, Karolinska Hospital CMM L8:00,

171 76 Stockholm, Sweden

e-mail: Dalila.Belazi@ki.se

P. Sjövall

Department of Chemistry and Materials Technology,

SP Technical Research Institute of Sweden,

PO Box 857, 50115 Borås, Sweden

106 Histochem Cell Biol (2009) 132:105–115

123

(Korn 1967). Lewis bases (electron donor, L) such as ter-

tiary amines and pyridines increase the reaction rate via the

formation of adduct OsO4L, which adds more rapidly to the

alkene (Thompson 2007). For example, the formation of

complexes involving osmium mono-esters and electron-

donor ligands from amino acids (Fig. 1d) (Nielson and

GriYth 1979) has been reported. Other osmium ester struc-

tures may involve the formation of a �-oxo bridge linking

osmium atoms (Fig. 1c).

The gradual blackening of tissue Wxed in OsO4, espe-

cially during the dehydration step through ethanol, is a

commonly observed phenomenon. The OsO4 binding to the

double bonds alone does not give rise to a color switch in

the stained tissue; the characteristic orange/brown/black

“osmium blacks” color is instead generated by the products

from the osmium ester reduction, forming a high accumula-

tion of osmium in the stained tissue region.

The Wrst step in the staining processes may result in the

formation of an osmium (VI) mono- or di-ester (Collin and

GriYth 1974). Observations that OsO4 stains unsaturated but

not saturated lipids make the hydrophobic fatty acid double

bond the most likely site for the osmium ester formation

(Hayes et al. 1963), while the hydrophilic carboxyl group

seems less likely. Furthermore, OsO4 has been found to react

in a 1:1 ratio to each oleWnic double bond, with unsaturated

tissue lipids and phospholipids, which suggests the forma-

tion of mono-esters (Baker 1958). However, there are spec-

troscopic evidences suggesting that osmium esters may

consist of a dimeric structure, cross-linking two separate

fatty acid chains (Collin and GriYth 1974). This may clarify

the role of OsO4 in the Wxation of membrane lipids, since

cross-linking of erstwhile double bonds via an Os2O2 bridge

would be expected to enhance the coherence of the lipid

structure (Wigglesworth et al. 1957). Thus, the formation of

mono- or di-esters must in part depend on the original dispo-

sition of double bonds and whether the chains of which these

bonds form a part can move relative to each other.

Finally, it has also been proposed that OsO4 simply

attaches to aliphatic side chains and proteins by hydrogen

bonds in the tissue (Litman and Barrnett 1972). However,

Raman, IR and Proton Magnetic Resonance show no evi-

dence for hydrogen bonding to OsO4, which may be due to

the considerable electron delocalization towards the central

osmium(VIII) atom (Collin and GriYth 1974).

Time-of-Xight secondary ion mass spectrometry (ToF-

SIMS) is a surface analytical method, which due to recent

instrumentation developments has emerged as a promising

method for chemical microanalysis of biological samples

(Belu et al. 2003; Winograd 2005; Johansson 2006).

BrieXy, mass spectra are recorded from the outermost

molecular layers of a solid sample, such as a biological cell

or tissue sample, by scanning a pulsed, focused, high-

energy (primary) ion beam across the sample surface. The

collision of the primary ions with the surface causes the

emission of atoms, molecular fragments and intact mole-

cules, some of which as (secondary) ions, from the surface

into vacuum. The secondary ions are extracted into a time-

of-Xight mass analyzer, providing mass spectra from each

raster point interrogated by the primary ion beam. The

recorded data can then be used to make ion images, dis-

playing the signal intensities from selected ions across the

analyzed surface area, or mass spectra from selected

regions of interest. Ideally, the method is capable of provid-

ing chemical images at a lateral resolution down to <0.5 �m

for speciWc molecules up to a few 1,000 Da, e.g., including

lipids in cell and tissue samples, without the need for chemi-

cal labeling or the application of a matrix to the sample.

In the present work, mass spectrometric imaging by

ToF-SIMS was used to detect and independently map the

spatial distributions of diVerent speciWc lipid species and

osmium oxide species in mouse adipose tissue. Comparison

of the lipid distributions before and after osmium staining

suggests that one type of osmium oxide binds speciWcally

to glyceride lipids [mainly triglycerides (TAG)] containing

Fig. 1 Binding of osmium to

oleWnic sites of unsaturated fatty

acids. a osmium mono-ester

formation, b osmium di-ester

formation and subsequent osmi-

um trioxide release, c osmium

di-ester involving an oxo bridge,

d osmium mono-ester coordinat-

ing with electron-donor groups

from amino acids (Korn

et al.1967; Nielson et al.1979)

Histochem Cell Biol (2009) 132:105–115 107

123

unsaturated C18 fatty acids, while another type of osmium

oxide, with a lower Os/O ratio, shows particle-like

accumulations without direct colocalization with the glyc-

eride lipids.

Materials and methods

Osmium tetroxide preparation

All glass material, including a 0.25 g OsO4 ampoule, a

probet and a glass rod, were carefully cleaned with concen-

trated HNO3 and rinsed in distilled water prior to

preparation of the OsO4 solution. Due to the toxicity of

OsO4, the wastewaters as well as the HNO3 used in the

cleaning process was stored in special bottles and further

eliminated in a controlled way.

A 4% osmium tetroxide aqueous solution was prepared

by adding the contents of the 0.25 g OsO4 ampoule into a

round-bottomed Xask containing 6.25 ml of distilled water.

The mixture was kept at room temperature for a few hours

until the OsO4 crystals were totally dissolved; 25 ml of 1%

OsO4 + PBS solution was then prepared by mixing 6.25 ml

of the 4% OsO4 aqueous solution with 18.75 ml PBS solu-

tion. The composition of the PBS solution was 0.01 M

phosphate buVer, 0.0027 M potassium chloride and

0.137 M sodium chloride. Finally, 30 ml of 0.33% OsO4

solution was prepared by mixing 10 ml of the 1% OsO4

solution with 20 ml of PBS solution.

Mice handling

Mice (C57BL) were brieXy anesthetized with CO2 and

decapitated. The gluteal fat pads were collected, immedi-

ately frozen on dry ice and stored at ¡80°C. The tissues

were further sectioned using a cryostat device (Leica Jung

CM 3000) at approximately ¡30°C. The 15-mm thin sec-

tions were placed on two diVerent pre-cooled glass slides.

The sectioned tissue was Wxed to the glass slides by gently

warming the opposite side of the slide with the Wnger. The

samples, properly stuck to the glass, were immediately

refrozen at ¡80°C.

Staining

Three samples were prepared for ToF-SIMS analysis and

taken out of the freezer before proceeding with the ToF-

SIMS analysis. Sample 1 (unstained) was used as a refer-

ence sample for comparison and was not stained. Sample 2

was thawed, and covered with 1 ml of 0.33% OsO4 solu-

tion. The sample was incubated 1 min at room temperature

before washing six times in distilled water. The washed

sample was then dried for 30 min before proceeding with

the analysis. Sample 3 was thawed and covered with 1 ml

of 1.0% OsO4 solution, incubated for 1 min, then washed in

distilled water six times and Wnally sequentially dehydrated

in 50, 70, 80 and 95% alcohol solutions, respectively.

TOF-SIMS analysis

The TOF-SIMS analysis was carried out using a TOF-

SIMS IV (ION-TOF, GmbH) instrument equipped with a

reXectron-type time-of-Xight mass spectrometer and a bis-

muth cluster primary ion source. The data were acquired

using 25 keV Bi3+ primary ions and low energy electron

Xooding for charge compensation, with the instrument opti-

mized for high mass resolution (bunched mode mass reso-

lution m/�m »5,000, lateral resolution 3–5 �m) or for high

image resolution (burst alignment mode mass resolution

m/�m »300, lateral resolution 200 nm). The pulsed primary

ion current was 0.1 pA in the bunched mode and 0.04 pA in

the burst alignment mode. The accumulated ion dose was

always kept below 1 £ 1012 ions/cm2. During analysis the

primary ion beam was scanned over the analysis area, col-

lecting separate mass spectra of the emitted secondary ions

from 128 £ 128 raster points (pixels). The recorded data

were used to produce total area mass spectra and/or ion

images showing the signal intensity distribution over the

analysis area of selected secondary ion peaks representing

speciWc compounds.

Results

Time-of-Xight secondary ion mass spectrometry of nega-

tive ions obtained on adipose tissue sections at diVerent

stages of the osmium staining procedure are shown in

Fig. 2. IdentiWcation of peaks in these spectra provides

detailed information about the chemical composition of the

tissue surface at the diVerent preparation stages. A selection

of peaks observed in the present study and their assign-

ments are listed in Table 1.

In the low mass region (m/z <100), the spectrum

obtained prior to osmium tetroxide incubation (Fig. 2a)

shows strong intensities from fragment peaks that can be

assigned to proteins (CN¡ and CNO¡), phospholipids

(PO2¡ and PO3

¡) and fatty acid fragments (CxHyO2¡). The

fragment ion assignments to proteins and phospholipids are

based on previous results from stained tissue (Sjövall et al.

2004) and from analysis of pure phospholipid (showing

high yields for PO2¡ and PO3

¡ and low yields for CN¡ and

CNO¡, see for example (Prinz et al. 2007)) and protein/

peptide (showing high yields for CN¡ and CNO¡ and low

yields for PO2¡ and PO3

¡) samples. After osmium incuba-

tion, the peaks corresponding to proteins and phospholipids

decrease to a very low signal level, indicating that a large

108 Histochem Cell Biol (2009) 132:105–115

123

part of these substances are removed from the surface dur-

ing osmium treatment, while the peaks corresponding to

fatty acid fragments remain strong.

In the mass region m/z 200–300, strong peaks from

diVerent fatty acids can be observed at all stages of prepara-

tion. Saturated and unsaturated fatty acids of diVerent chain

lengths (labeled as CX:Y, where X denotes the number of

carbon atoms and Y is the number of double bonds) are

readily separated in the spectra, see inset in Fig. 2b, which

means that the speciWc fatty acids can be independently

monitored. In the high mass region of the spectrum

obtained before osmium treatment (Fig. 2a), molecular ions

of various triglycerides (TAG) and phosphatidylinositol

(PI) can be observed. The PI peak is not present after

osmium treatment, consistent with removal of phospholip-

ids during this step (as suggested above), while the TAG

Fig. 2 Negative TOF-SIMS

spectra from adipose tissue a

before Os staining, b after Os

staining and c after subsequent

ethanol rinsing. Representative

peaks for proteins (pr), phos-

phate (ph), C16 and C18 fatty

acids, triglycerides (TAG), phos-

phatidylinositol (PI) and diVer-

ent OsxOy¡ ions are indicated

a)

b)

c)

Table 1 Observed ions in TOF-SIMS spectra from adipose tissue and their chemical assignments

For the fatty acids, DAGs and TAGs, only the saturated ions are included. Corresponding unsaturated ions were observed at the expected m/z val-

ues (subtract m/z 2.016 for each included double bond)

Ion Observed

mass (Da)

Theoretical

mass (Da)

Mass accuracy

(ppm)

Assignment

CN¡ (Sjövall et al. 2004) 26.004 26.003 65.8 Protein fragment

CNO¡ (Sjövall et al. 2004) 42.000 41.998 47.7 Protein fragment

PO2¡ (Sjövall et al. 2004) 62.961 62.963 ¡39.9 Phospholipid fragment

PO3¡ 78.958 78.958 ¡9.2 Phospholipid fragment

C16H31O2¡ (palmitate) 255.216 255.233 ¡62.7 C16:0 fatty acid

C18H35O2¡ (stearate) 283.260 283.264 ¡14.5 C18:0 fatty acid

C51H97O6¡ (Sjövall et al. 2008) 805.751 805.727 27.3 TAG 48:0

C53H101O6¡ (Sjövall et al. 2008) 833.776 833.761 19.9 TAG 50:0

C55H105O6¡ (Sjövall et al. 2008) 861.771 861.792 ¡22.2 TAG 52:0

C47H82PO13+ (Sjövall et al. 2004) 885.603 885.551 59.4 PI38:4

OsO3¡ 239.932 239.947 ¡56.3 Osmium oxide (I)

Os2O4¡ 445.880 445.900 ¡44.8 Osmium oxide (II)

C5H15NPO4+ (Pacholski et al. 1998;

Prinz et al. 2007)

184.074 184.074 2.2 Phosphocholine

C33H63O4+ (Sjövall et al. 2008) 523.468 523.473 ¡8.8 DAG30:0 (¡OH)

C35H67O4+ (Sjövall et al. 2008) 551.500 551.503 ¡8.9 DAG32:0 (¡OH)

C37H71O4+ (Sjövall et al. 2008) 579.535 579.535 1.4 DAG34:0 (¡OH)

C39H75O4+ (Sjövall et al. 2008) 607.571 607.568 5.5 DAG36:0 (¡OH)

Histochem Cell Biol (2009) 132:105–115 109

123

peaks are present in the spectra after osmium incubation

and, at a weaker intensity, after ethanol dehydration.

After osmium tetroxide incubation, a large number of

peaks appear in the spectra that can be identiWed as diVerent

OsxOy¡ ions. The most intense of these peaks are observed

at m/z 234–240, corresponding to the diVerent natural iso-

topes of OsO3¡ (see Fig. 3a). Comparison with the theoreti-

cal isotope pattern reveals that there is also a contribution in

the spectrum from OsO3H¡ (Fig. 3a–c). At higher masses,

evenly distributed groups of peaks are observed, at a sepa-

ration that approximately corresponds to the mass of OsO.

These peaks can be identiWed as osmium oxide cluster ions

of the type OsO3(OsO)x¡, with adjoining peaks correspond-

ing to clusters with one less or one more oxygen atom.

Figure 3d shows the observed isotope pattern from one

such cluster, Os3O5¡, which by comparison with the theo-

retical isotope pattern (Fig. 3e–f) suggests contribution

from Os3O5H¡ clusters. The isotopic pattern and the mass

accuracy (see Table 1) observed for the osmium oxide

peaks provide strong evidence for the correct assignment of

these peaks. The observation of clusters of the type

OsO3(OsO) in the spectrum suggests the formation of bulk

oxides in the tissue sample. However, it is not possible

from the ToF-SIMS data to determine the exact chemical

identity of the osmium oxide or osmium oxide-containing

species in the tissue sample.

We did not detect any peaks that could conclusively be

assigned to vicinal diols, although their presence in the tis-

sue sample is highly expected after OsO4 staining. The

most likely reason for this is that these compounds may be

easily fragmented during the ion bombardment process or

that they may not form stable ions at suYcient yields to be

detected in our measurement.

Figure 4 shows ion images obtained prior to osmium

incubation at two diVerent magniWcations. The images

show the signal intensity distributions of the indicated ions

over the analysis area, (brighter pixels indicate higher sig-

nal intensity), reXecting the spatial distribution of the corre-

sponding compound on the tissue surface. The upper row

shows 500 �m £ 500 �m images (at high mass resolu-

tion—low image resolution) representing the spatial distri-

bution of protein fragments (CNO), phospholipids (PO2)

and the C18:0 (saturated) and C18:1 (unsaturated) fatty

acids on the adipose tissue sample. The images in the lower

row are magniWed high resolution images obtained from the

area indicated by the square in the upper CNO image. Com-

paring the images in the lower row, it can be seen that the

protein and phospholipid distributions are similar, although

with some minor diVerences, while the distributions of the

fatty acids are complementary to that in the phospholipid

image. Furthermore, the distributions of the saturated and

unsaturated fatty acids are very similar in these images.

Fig. 3 Comparison between observed and theoretical isotope pat-

terns for two osmium oxide ions, OsO3¡ and Os2O5

¡. The theoretical

isotope pattern in b was calculated by adding contributions from

OsO3¡ and OsO3H¡ at a fraction determined by the observed signal

intensities at m/z 240 and 241 (which can be assumed to provide

contributions from only OsO3¡ and OsO3H¡, respectively). The the-

oretical isotope pattern for OsO3¡ only is shown in (c). d–f Shows

the equivalent comparison for the peaks assigned to Os3O5¡ and

Os3O5H¡. The peaks labeled with (asterisk) in (a) originate from

unassigned organic fragments

110 Histochem Cell Biol (2009) 132:105–115

123

However, regions were also observed in which these distri-

butions were not entirely similar (not shown). A possible

interpretation of these images is that the protein and

phospholipid-rich regions correspond to cell membrane

structures on the tissue section surface, while the fatty acid-

rich regions correspond to glyceride deposits.

Ion images showing the spatial distributions of two

diVerent osmium oxide ions, phosphocholine and several

fatty acid ions after osmium staining (before ethanol rins-

ing) are presented in Fig. 5. The two osmium oxide images

show striking diVerences suggesting that there are diVerent

forms of osmium oxide present in the tissue sample, one

which is distributed according to the OsO3¡ image (in the

following denoted type I) and another form with the parti-

cle-like distribution of the Os2O5¡ image (type II). Images

of all osmium clusters with two or more osmium atoms (see

Fig. 2c) show similar distributions as the Os2O5¡ image.

The particle structures of the Os2O5¡ image are seen also in

the total ion image, which indicates that they have a topo-

graphic particle structure, while the distribution of the

OsO3¡-related osmium oxide does not show any topo-

graphic features. The phosphocholine image shows that the

small amounts of phosphatidylcholine and/or sphingomye-

lin remaining in the tissue sample after osmium staining is

strongly localized to the type II osmium oxide. Also the

fragment ion assigned to proteins (CN¡ and CNO¡) were

strongly localized to the type II osmium oxide (not shown).

The unsaturated C18 fatty acid (C18:1 and C18:2)

images in Fig. 5 show striking similarities in spatial distri-

butions with that of the type I osmium oxide ion (OsO3¡).

In contrast, the saturated fatty acid images (C18:0 and

C16:0) show complementary localizations to the type I

osmium oxide. Furthermore, the unsaturated C16 ions

(C16:1 and C16:2) show distributions that are more similar

to the saturated fatty acid images than to the unsaturated

C18 fatty acid and type I osmium oxide images. Ion images

of C14 fatty acids (not shown) display similar features as

the C16 fatty acids. The complementary localization

between the unsaturated and saturated C18 fatty acids and

the colocalization between the unsaturated C18 fatty acids

and the type I osmium oxide are further demonstrated in the

two three-color overlay images in Fig. 5. In the upper of

these three-color overlay images, showing distinctive red

and blue areas representing unsaturated and saturated fatty

acids, respectively, it can also be seen that neither of the

unsaturated or saturated fatty acids are localized to the par-

ticle structures of the type II osmium oxide (green). Frag-

mentation of glycerides mainly TAG:s, the main lipid

constituent of the adipose tissue (Body 1988), is likely to

produce a release of fatty acid ions that are colocalized with

Fig. 4 Negative ion images from adipose tissue before Os staining

showing the signal intensity distributions of CNO¡, representing pro-

teins, PO2¡, representing phospholipids and two speciWc fatty acids,

C18:0 and C18:1. The upper row shows low resolution images at a

Weld of view of 500 �m £ 500 �m and the lower row shows high res-

olution images from the 128 �m £ 128 �m area indicated by the gray

square. Brighter pixels means higher signal intensities, as shown by

the color scales. Each image is normalized to the maximum signal

intensity in a single pixel. “mc” stands for maximum number of detect-

ed ions (counts) in a single pixel and “tc” stands for total counts in the

entire image. The small (2.4%) signal intensity contribution from

C18:1 to the C18:0 signal intensity was not taken into account

Histochem Cell Biol (2009) 132:105–115 111

123

their TAG:s. These images strongly suggest that the

osmium oxide (type I) speciWcally binds to unsaturated C18

fatty acids and to TAG containing unsaturated C18 and not

to saturated fatty acids or unsaturated C16 (or C14) fatty

acids and TAG:s containing them. The strong complemen-

tary localization between saturated and unsaturated C18

fatty acids was not seen prior to osmium staining (Fig. 4),

which suggests that a certain redistribution of lipids may

occur in the tissue sample during the osmium staining pro-

cedure

Figure 6 shows positive ion spectra from the adipose tis-

sue sections at the three diVerent stages of preparation. A

number of peaks identiWed as monoacylglycerides (MAG)

and diacylglycerides (DAG) are observed at all three prepa-

ration stages. DAG ions representing a variety of carbon

chain lengths and number of double bonds are observed

(see Fig. 6a; Table 1). For this particular study with adipose

tissue, where TAG is the major constituent (component)

(Body 1988), we attribute the main contributions to the

fatty acid, MAG and DAG peaks to fragmentation of TAG.

However, contributions from free fatty acids and DAG lip-

ids may also be present. In the spectrum prior to osmium

treatment, peaks corresponding to phosphocholine, the

head group of phosphatidylcholine and sphingomyelin, are

observed at m/z 166 and 184. However, these peaks

decrease dramatically in intensity after osmium treatment,

consistent with the decrease of the phosphate peaks in the

negative spectra (Fig. 2), and is attributed to a loss of phos-

pholipids from the tissue sample during osmium staining.

Ion images reXecting the spatial distributions of a variety

of DAG ions after osmium incubation are shown in Fig. 7.

The images are consistent with DAG distributions obtained

by adding the signal intensities of the fatty acid components

in the DAG ions. For example, comparing the images of

DAG ions containing 36 carbon atoms in the tail groups but

diVerent number of double bonds (36:X, X = 0–3), it can be

Fig. 5 Ion images from adipose tissue after incubation in osmium

tetroxide (no ethanol rinsing) showing the lateral distribution of two

diVerent osmium oxide ions, phosphocholine and several fatty acids

ions. The fatty acid labels (CX:Y) speciWes the number of carbon

atoms (X) and double bonds (Y) in the ion. The C16:0 image was

obtained using the isotope peak at m/z 256, due to signal saturation in

the major isotope peak at m/z 255. The three-color images are overlays

of the indicated images in blue, red and green. Purple indicates areas

of overlap of blue and red (C18:2 and OsO3) and yellow indicates over-

lap of red and green (OsO3 and Os2O5)

112 Histochem Cell Biol (2009) 132:105–115

123

observed that the distribution of the saturated DAG (36:0)

is similar to the C18:0 fatty acid distribution (see bottom

row of Fig. 7), while the distributions of the 36:2 and 36:3

DAG ions show the same spatial distribution features as the

unsaturated C18 fatty acids. The complementary localiza-

tion of 36:0 and 36:3 is highlighted in the red/green overlay

image in Fig. 7. Furthermore, the 36:1 image displays a

largely homogenous distribution, which is very similar to

the one that is obtained by adding the signal intensities of

the C18:0 and C18:1 fatty acid ion images (see bottom row

in Fig. 7). Similar observations can be made for the 34:X

DAG ions, which can be expected to mainly contain one

C16 and one C18 fatty acid, while for the 32:X DAG ions,

which can be expected to contain mainly C16 fatty acids,

the features of the unsaturated C18 fatty acid distributions

are not observed (however diYcult to see due to the low

signal intensities).

The ion images in Fig. 8 shows that the distributions of

osmium oxide and unsaturated fatty acids are changed dur-

ing ethanol rinsing of the osmium-stained tissue sections.

The two types of osmium oxide ions and the unsaturated

C18 fatty acid ions (C18:1 and C18:2) all show homoge-

nous distributions over the entire tissue sample surface

after ethanol rinse (the structure seen in the images can be

attributed to topographic eVects, based on their similarity

with the total ion image (not shown). The saturated fatty

acids and the unsaturated C16 fatty acid ions show inhomo-

genous distributions similar in structure to the ones

observed prior to ethanol rinsing. The colocalization of the

osmium oxide to C18:1 and C18:2 and not to C16:1 (or the

saturated fatty acids) conWrms the suggestion made above

that osmium oxide selectively binds to lipids containing

unsaturated C18 fatty acids (and not to unsaturated C16).

A possible explanation to the change in spatial distribu-

tions during ethanol rinsing is that both the type II osmium

oxide and complexes consisting of unsaturated C18 fatty

acids and type I osmium oxide may be dissolved in the

ethanol and subsequently homogenously distributed on the

tissue surface during drying.

Summary and discussion

The results from this work provide detailed information

about the localization of osmium oxides and speciWc lipids

in adipose tissue upon staining by OsO4. The main results

can be summarized as follows:

– Before osmium incubation, unsaturated and saturated fatty

acid, diglyceride and triglyceride ions are mainly colocal-

ized in speciWc structures, most likely corresponding to tri-

glycerides deposits in the adipose tissue. Complementary

localized structures, possibly corresponding to cell mem-

brane regions, show strong signal intensity from ions char-

acteristic for phospholipids and proteins.

– After osmium staining, two forms of osmium oxide were

observed on the tissue surface, showing diVerent spatial

distributions and diVerent Os/O ratios. Type I osmium

oxide shows an inhomogenous distribution without topo-

graphic features and was monitored by the OsO3¡ ion.

Type II osmium oxide shows a particle-like distribution

with topographic features and was monitored by ion

clusters of the form OsO3(OsO)n¡, suggesting a less oxy-

gen-rich oxide than the type I osmium oxide.

– Unsaturated C18 fatty acids (C18:1 and C18:2), and

diglycerides containing these fatty acids, were found

to be colocalized with type I osmium oxide, while the

Fig. 6 Positive TOF-SIMS

spectra from adipose tissue a

before Os staining, b after Os

staining and c after subsequent

ethanol rinsing. Peaks from

diVerent diacylglyceride (DAG)

and monoacylglyceride (MAG)

ions, and the phosphocholine

(pc) ion, representing the head

group of phosphatidylcholine or

sphingomyelin, are indicated.

The inset shows an expanded

region of the spectrum highlight-

ing peaks from diVerent DAG

ions

a)

b)

c)

Histochem Cell Biol (2009) 132:105–115 113

123

Fig. 7 Ion images from adipose tissue after staining with osmium

tetroxide (no ethanol rinsing) showing the lateral distributions of

diVerent DAG ions indicated by the X:Y notation (X total number of

carbon atoms, Y total number of double bonds). The bottom row shows

the C18:0 fatty acid image and the sum of the indicated fatty acid

images. The red/green image demonstrates the complementary local-

ization of the saturated DAG and the DAG with three double bonds

114 Histochem Cell Biol (2009) 132:105–115

123

saturated C18 fatty acid and C16 (saturated and unsatu-

rated) fatty acids, and diglycerides containing these fatty

acids, show complementary localizations as compared to

the type I oxide.

– Fragment peaks assigned to proteins and phospholipids

showed a strong decrease in signal intensity after OsO4

incubation, indicating that a considerable fraction of the

proteins and phospholipids were removed from the tissue

surface during the OsO4 incubation step. However, the

remaining proteins and phospholipids were strongly

colocalized with type II osmium oxide.

– After ethanol dehydration of the OsO4-stained tissue,

both type I and type II osmium oxides, as well as the

unsaturated C18 fatty acids, show homogeneous distri-

bution covering the entire tissue surface. The saturated

C18 and the C16 (saturated and unsaturated) fatty acids

chains show more inhomogenous distributions.

The observed colocalization between type I osmium oxide

and unsaturated C18 fatty acids, and glyceride lipids con-

taining these fatty acids, is consistent with the established

view that OsO4 binds to unsaturated lipids. However, our

results also indicate that the osmium oxide binding is con-

siderably stronger to unsaturated C18 fatty acids chains, as

compared to the binding to unsaturated C16 (or C14) fatty

acids. This distinction between unsaturated C18 and C16

fatty acids was a highly unexpected observation but, none-

theless, quite evident from our data. To our knowledge,

there are no published results that can either conWrm or

contradict our observation. Additional measurements, using

ToF-SIMS or complementary techniques, are therefore

needed in order to verify the observation and to determine

how general the eVect is. It is, e.g., possible that the pre-

ferred binding of osmium oxide to C18 fatty acids is

speciWc to adipose tissue, in which the lipid content is

dominated by triglycerides, while the binding may be

diVerent in other tissue types, in which phospholipids

constitute a larger fraction of the lipid content.

The observed colocalization between type II osmium

oxide and proteins/phospholipids is consistent with the

reported binding of osmium oxide to proteins and ali-

phatic phospholipid side chains. The reduced nature of

type II osmium oxide may be due to reducing activity of

electron donor groups located in proteins, which may have

reduced higher oxidation state osmium oxides. In addi-

tion, our results show that a large fraction of the proteins

and phospholipids originally in the tissue is extracted out

of from the tissue during the OsO4 incubation and staining

steps. As described in Horobin (1982), it is quite common

that tissue material are lost as an eVect of Wxation and

staining. The proteins and phospholipids lost during the

diVerent phases of the Wxation/staining procedure, could

be found mainly in the Wxative solution while the rest

were lost in the water washes and possibly in other pro-

cessing Xuids (Horobin 1982). Same observation have

been made regarding the losses of phospholipids in a

study done by Takahashi et al. (1989) involving osmium

tetroxide post Wxation (and losses of phospholipids during

the procedure).

Fig. 8 Ion images from adipose tissue after staining with osmium

tetroxide and subsequent ethanol rinsing. The OsO3¡ image was

obtained using the isotope peak at m/z 239, due to signal saturation in

the major isotope peak at m/z 240. The purple color in the red/green/

blue overlay image indicates colocalization of osmium oxide and the

unsaturated fatty acid C18:1

Histochem Cell Biol (2009) 132:105–115 115

123

Considering the reported reaction mechanism for OsO4

staining in cells and tissues, involving initial formation of

an osmium ester and subsequent reduction resulting in Os

accumulations in the tissue (see Fig. 1), a possible interpre-

tation of our results is that type I osmium oxide is related to

and therefore monitors osmium esters formation, while type

II monitors the reduced osmium species. This interpretation

is supported by the observed particle-like structure of type

II osmium oxide, resembling solid deposits of reduced Os.

Finally, the diVerent images show a strong complemen-

tary localization between saturated and unsaturated C18

fatty acids after OsO4 incubation, while such complemen-

tarity was not observed prior to OsO4 incubation. This sug-

gests that the OsO4 incubation may give rise to some

redistribution of lipids in the tissue, at least at the »10 �m

length scale. However, the macroscopic distribution of

osmium oxide found in the tissue after OsO4 staining is still

likely to reXect the original distribution of unsaturated

(C18) fatty acids and their associated glycerides. Further-

more, our results, indicating the redistribution of lipids in

triglyceride deposits in adipose tissue during OsO4 staining,

does not necessarily mean that similar redistributions occur

in other types of tissues. Additional studies are needed in

order to determine whether lipid redistributions may be a

problem also in other tissue types.

The study of the structure and the contents of tissues and

cells by light or electron microscopy usually necessitates

staining and Wxation. This is mainly done in order to protect

form and content of the native tissue from diVerent kind of

degradation during the processing. DiVerent Wxatives can

give diVerent chemical modiWcations, leaving the tissue in

diVerent physical states (Horobin 1982). The use of osmium

tetroxide as a staining dye in this study suggests that chemi-

cal and physical modiWcations are possible.

With its combination of chemical analysis and imaging,

and an ability to work with tissue sections similar or identi-

cal to those in histology, ToF-SIMS is highly useful for elu-

cidating the chemical background of the histological

staining techniques. ToF-SIMS also allows the study of

biological tissues without the need of any kind of Wxation

or staining, keeping the tissue intact from possible degrada-

tion resulting from chemical or physical alteration due to

the preparation procedure. With a more widespread use and

an increased experience from this type of studies, TOF-

SIMS may become an important complement to conven-

tional microscopy when detailed chemical information

about a tissue sample is needed.

Acknowledgments This research was supported by EC FP6 funding

(contract no.005045 NANOBIOMAPS), Swedish Governmental

Agency for Innovation systems (VINNOVA), Swedish Research

Council (VR), Swedish Heart and Lung Foundation (20065011), Fun-

dació la Marató de TV3 and an unconditional grant from Baxter

Healthcare.

References

Alberts B, Wilson J, Johnson A, Lewis J, RaV M, Roberts K (2007)

Visualizing cells. In: Molecular biology of the cell, 5th edn.

Garland Science, New York, 579–616

Baker JR (1958) Fixation in cytochemistry and electron-microscopy.

J Histochem Cytochem 6:303–308

Belu AM, Graham DJ, Castner DG (2003) Time-of-Xight secondary

ion mass spectrometry: techniques and applications for the char-

acterization of biomaterial surfaces. Biomaterials 21:3635–3653

Body DR (1988) The lipid composition of adipose tissue. Prog Lipid

Res 27:39–60

Bozzola JJ, Russell DL (1999) Specimen preparation for transmission

electron microscopy. In: Electron microscopy, 2nd edn. Jones and

Bartlett Publishers, Inc., Sudbury, pp 16–47

Collin RJ, GriYth WP (1974) Mechanism of tissue component staining

by osmium tetroxide. J Histochem Cytochem 22:992–996

Hayes TL, Lindgren FT, Gofman JW (1963) Quantitative determina-

tion of the osmium tetroxide–lipoprotein interaction. J Cell Biol

19(1):251–255

Horobin RW (1982) Fixation and other loss-limiting procedures.

In: Histochemistry. Butterworths, London, pp 19–55

Johansson B (2006) ToF-SIMS imaging of lipids in cell membranes.

Surf Interface Anal 38:1401–1412

Korn ED (1967) A chromatographic and spectrophotometric study of

the reaction of osmium tetroxide with unsaturated lipids. J Cell

Biol 34:627–638

Litman RB, Barrnett RJ (1972) The mechanism of the Wxation of tissue

components by osmium tetroxide via hydrogen bonding. J Ultra-

struct Res 38(1):63–86

Moore JE, Zouridakis G (2003) Biomedical technology and devices

handbook. CRC, Boca Raton, p 164

Nielson AJ, GriYth WP (1979) Tissue Wxation by osmium tetroxide. A

possible role for proteins. J Histochem Cytochem 27(5):997–999

Pacholski ML, Cannon DM Jr, Ewing AG, Winograd N (1998) Static

time-of-Xight secondary ion mass spectrometry imaging of

freeze-fractured, frozen-hydrated biological membranes. Rapid

Commun Mass Spectrom 12(18):1232–1235

Prinz C, Höök F, Malm J, Sjövall P (2007) Structural eVects in the

analysis of supported lipid bilayers by time of Xight secondary ion

mass spectrometry. Langmuir 23(15):8035–8041

Sjövall P, Lausmaa J, Johansson B (2004) Mass spectrometric imaging

of lipids in brain tissue. Anal Chem 76(15):4271–4278

Sjövall P, Johansson B, Belazi D, Stenvinkel P, Lindholm B, Lausmaa J,

Schalling M (2008) TOF-SIMS analysis of adipose tissue from

patients with chronic kidney disease. Appl Surf Sci 255(4):

1177–1180

Studer D, Humbel BM, Chiquet M (2008) Electron microscopy of

high pressure frozen samples: bridging the gap between cellular

ultrastructure and atomic resolution. Histochem Cell Biol 130:

877–889

Takahashi T, Okuda Y, Kohno T et al (1989) Fixation eVect of phos-

pholipids using tannic acid. Part 1: artiWcial lung surfacant.

Kanagawa Shigaku 23(4):610–621

Thompson M (2007) Osmium tetroxide (OsO4) http://www.chm.bris.

ac.uk/motm/oso4/oso4h.htm (visited Dec-2007)

Wigglesworth HB (1957) The use of osmium in the Wxation and stain-

ing of tissues. Proc R Soc Lond Biol 147:185–199

Winograd N (2005) The magic of cluster SIMS. Anal Chem 77(7):

142A–149A

Zollinger H (2003) Dyes in biochemistry, biology, medicine and analyt-

ical chemistry. In: Color chemistry, 3rd edn, Wiley-VCH, Zürich,

pp 543–546