Journal of Chromatography B, 1047 (2017) 131–140
Contents lists available at ScienceDirect
Journal of Chromatography B
jou rnal homepage: www.elsevier.com/locate/chromb
MALDI imaging mass spectrometry analysis—A new approach for
protein mapping in multiple sclerosis brain lesions
a,b,1 a,1 c
Giuseppina Maccarrone , Sandra Nischwitz , Sören-Oliver Deininger ,
a d,e d
Joachim Hornung , Fatima Barbara König , Christine Stadelmann ,
b,1 a,f,∗,1
Christoph W. Turck , Frank Weber
a
Max Planck Institute of Psychiatry, Kraepelinstr. 2-10, 80804 Munich, Germany
b
Department of Translational Research in Psychiatry, Max Planck Institute of Psychiatry, Germany
c
Bruker Daltonik GmbH, Fahrenheitstr. 4, 28359 Bremen, Germany
d
Institute of Neuropathology, University Medical Center Göttingen, Robert-Koch-Str. 40, 37075 Göttingen, Germany
e
Institut für Pathologie, Klinikum Kassel, Mönchebergstr. 41-43, 34125 Kassel, Germany
f
Medical Park Bad Camberg, Obertorstr. 100-102, 65520 Bad Camberg, Germany
a r t i c l e i n f o a b s t r a c t
Article history: Multiple sclerosis is a disease of the central nervous system characterized by recurrent inflammatory
Received 21 February 2016
demyelinating lesions in the early disease stage. Lesion formation and mechanisms leading to lesion
Accepted 1 July 2016
remyelination are not fully understood. Matrix Assisted Laser Desorption Ionisation Mass Spectrom-
Available online 1 July 2016
etry imaging (MALDI–IMS) is a technology which analyses proteins and peptides in tissue, preserves
their spatial localization, and generates molecular maps within the tissue section. In a pilot study we
Keywords:
employed MALDI imaging mass spectrometry to profile and identify peptides and proteins expressed in
MALDI imaging mass spectrometry
normal-appearing white matter, grey matter and multiple sclerosis brain lesions with different extents
LC–ESI–MS/MS
of remyelination. The unsupervised clustering analysis of the mass spectra generated images which
Multiple sclerosis
Demyelination reflected the tissue section morphology in luxol fast blue stain and in myelin basic protein immunohis-
Remyelination tochemistry. Lesions with low remyelination extent were defined by compounds with molecular weight
Thymosin beta-4 smaller than 5300 Da, while more completely remyelinated lesions showed compounds with molecular
weights greater than 15,200 Da. An in-depth analysis of the mass spectra enabled the detection of cortical
lesions which were not seen by routine luxol fast blue histology. An ion mass, mainly distributed at the
rim of multiple sclerosis lesions, was identified by liquid chromatography and tandem mass spectrom-
etry as thymosin beta-4, a protein known to be involved in cell migration and in restorative processes.
The ion mass of thymosin beta-4 was profiled by MALDI imaging mass spectrometry in brain slides of
12 multiple sclerosis patients and validated by immunohistochemical analysis. In summary, our results
demonstrate the ability of the MALDI–IMS technology to map proteins within the brain parenchyma and
multiple sclerosis lesions and to identify potential markers involved in multiple sclerosis pathogenesis
and/or remyelination.
© 2016 Elsevier B.V. All rights reserved.
1. Introduction
Multiple sclerosis is a disease of the central nervous system
that is characterized by recurrent inflammatory, demyelinated
lesions—at least in the early stages. The mechanisms of lesion for-
mation and lesion repair, especially remyelination are not fully
Abbreviation: MALDI–IMS, matrix assisted laser desorption ionisation–imaging
mass spectrometry; LC–ESI–MSMS, liquid chromatography–electro spray ionisation understood.
tandem mass spectrometry; LFB, luxol fast blue; CNP-ase, 2 ,3 -cyclic nucleotide
In the early disease stages multiple sclerosis is characterised
3 -phosphodiesterase; PLP, proteolipid protein.
∗ by focal lympho- and monocytic infiltrations, in the later stages
Corresponding author at: Max Planck Institute of Psychiatry, Kraepelinstr. 2,
microglial activation and ongoing neurodegeneration prevail [1].
D-80804 Munich, Germany.
Remyelination which occurs to some extent in most multiple scle-
E-mail address: [email protected] (F. Weber).
1
These authors contributed equally to this work.
http://dx.doi.org/10.1016/j.jchromb.2016.07.001
1570-0232/© 2016 Elsevier B.V. All rights reserved.
132 G. Maccarrone et al. / J. Chromatogr. B 1047 (2017) 131–140
rosis lesions is an important mechanism that protects axons and peptides in human multiple sclerosis brain tissues including lesions
may prevent chronic disability [2,3]. The extent of remyelination with different extents of remyelination.
varies in different lesions of an individual multiple sclerosis patient.
The underlying mechanisms leading to this discrepancy and the 2. Materials and methods
lack of remyelination in many lesions are not fully understood
[4]. Several studies, however, suggest that successful regeneration 2.1. Brain tissue samples
depends on a signaling environment conducive to remyelina-
tion, which is provided in the context of acute inflammation [5]. The multiple sclerosis brain tissues were provided by the UK
Macrophages are mainly found in acute and chronic active multi- Multiple Sclerosis Tissue Bank, Division of Neuroscience and Men-
ple sclerosis lesions with ongoing inflammation and are involved tal Health Imperial College, London with fully informed consent
in demyelination but also in remyelination processes of the central from the donor and the next of kin. The brain tissue includes
nervous system [6,7]. Nerve growth factors, proteases, and other normal-appearing white matter, grey matter and lesions. The brain
proteins that affect myelin stability or have anti-oxidative activi- tissue was collected post-mortem with the post mortem time less
ties seem to play an important role in either promoting or inhibiting than 24 h. Cerebral hemispheres were coronally sliced, blocked and
remyelination [8–11]. However, the exact mechanisms of remyeli- frozen by immersion in cold liquid iso-pentane and then stored at
◦
nation and the proteins involved remain elusive. −75 C.
An ongoing disease process is reflected by qualitative and quan-
titative molecular alterations. In case of multiple sclerosis the
2.2. Discovery samples
changes of protein localisation and composition in the brain lesions
reflects the alteration of myelination and the effect of inflammation.
For the pilot study brain tissue slides from 2 brain blocks labelled
Thus molecular maps and relative abundance profiling of proteins
A and B of a 71-year-old female patient with secondary progressive
and protein fragments expressed in lesions with different extents
multiple sclerosis were used. The patient had initially a relapsing
of remyelination are of great value for an improved understanding
and remitting disease course for 20 years, then secondary progres-
of multiple sclerosis aetiology.
sive for about 14 years. A magnetic resonance imaging (MRI) scan of
Typically, in neuropathology protein analysis is performed by
the brain showed lesions suggestive of multi-focal demyelination.
immunohistochemistry and two-dimensional polyacrylamide gel
No other neurological disorders were noted. Diagnosis of multiple
electrophoresis [12,13]. Recently, mass spectrometry-based meth-
sclerosis was confirmed post-mortem by histopathology.
ods have been applied for the identification of lesion-specific
proteins of distinct histopathological types of multiple sclerosis
2.3. Validation samples
brain lesions [14]. Although this type of analysis results in com-
prehensive proteomic profiles, information about protein location,
For the MALDI Imaging and immunohistochemical validation
the main focus in neuropathological analyses, is not captured. An
studies brain samples from 12 multiple sclerosis patients were
important technological advance in protein analyses that addresses
used. Clinical characteristics and the death tissue-preservation
this shortcoming is Matrix Assisted Laser Desorption Ionisation
interval are summarized in Table 1.
Imaging Mass Spectrometry (MALDI–IMS). This technique yields
a tri-dimensional image of proteins and peptides capturing molec-
2.4. Histological analysis
ular masses, relative abundances and spatial coordinates [15–17].
MALDI-IMS is an unbiased approach to look for potential disease ◦
The brain tissue was sliced at −18 C into 10 m thick sec-
related proteins and peptides (complementing immunohistochem- ◦
tions using a microtome (Leica, CM30050), and stored at −20 C.
istry). In the present study our main focus was to investigate the
To identify lesions, standard hematoxylin and eosin and Luxol Fast
capability of MALDI–IMS to map proteins and peptides associated
Blue (LFB) staining (Sigma, Solvent Blue 38) were used to visualise
either with remyelination or resulting from demyelination pro-
myelin, and counterstaining with cresyl violet (Sigma, Cresyl-Violet
cesses which occur during the development of multiple sclerosis
Acetate) or hematoxylin and periodic acid-Schiff were performed.
lesions. We applied this new technology to profile proteins and
To determine the extent of remyelination in multiple sclerosis
lesions, LFB-stained brain sections were scanned with a Color View
Table 1
Clinical characteristics and death tissue-preservation interval of multiple sclerosis patients used in the validation studies.
Patient Age [years] Gender Disease duration Disease course MS specific treatment Cause of death Death-Tissue
[years] (except steroids) preservation
interval [h]
MS049 75 m 38 SPMS none aspiration pneumonia 8
MS055 47 f 32 SPMS none pneumonia 15
MS066 86 f more than 50 N/A none pneumonia 21
MS080 71 f 35 SPMS none bowel 24 blockage/post-operative
comlpications
MS094 42 f 6 PP none pneumonia 11
MS098 57 m 33 SPMS none sepsis secondary to urinary 21
tract infection
MS105 73 m N/A N/A N/A pneumonia 8
MS107 38 m 16 SPMS vitamin B12 injections, tricyclic pneumonia 19
anti-depressants and secret remedies
MS114 52 f 15 SPMS none Pneumonia, sepsis, 12
pulmonary embolism
MS136 40 m 9 SPMS interferone beta 1a Respiratory failure, sepsis 10
MS154 34 f 11 SPMS none pneumonia 12
MS249 59 f 42 SPMS none Chest infection 8
G. Maccarrone et al. / J. Chromatogr. B 1047 (2017) 131–140 133
digital camera (Soft Imaging System, Münster, Germany) mounted 2.7. Statistical analysis
on an Olympus BX51 microscope (Olympus, Tokyo, Japan). Total
white matter, total lesion area (including remyelinated parts of The statistical analyses and data processing of the MALDI spec-
the lesion) as well as remyelinated lesion areas were measured tra were performed using the ClinProTools 2.2 software package
®
using Analysis software (Soft Imaging System GmbH, Münster, (Bruker Daltonik, Germany). First the mass spectra were internally
Germany). The percentage of remyelinated lesion area was calcu- aligned on common peaks and normalised with respect to the total
lated by dividing the total lesion area by the remyelinated lesion ion counts. Then an average spectrum was created and used for the
area and then multiplying by 100. peak picking process. An unsupervised hierarchical statistical anal-
To determine the distribution of Proteolipid Protein (PLP) and ysis was performed as previously described [19]. Briefly, the data
2 ,3 -Cyclic Nucleotide 3 -Phosphodiesterase (CNPase), cryostat were first pre-processed by Principal Component Analysis and then
sections were air-dried, incubated for 1 h in chloroform, and sub- the spectra were clustered applying unit variance scaling, Euclidean
sequently fixed for 30 min in 4% paraformaldehyde and for epitope distance metric and Ward linkage method. The clustered mass
retrieval in 10 mM citric acid buffer (pH 6.0) in a steamer. Tis- spectra were illustrated by dendrogram with color-code branches
sue sections were incubated in phosphate buffered saline solution whose visualization generated a tissue MALDI image. The color-
◦
containing 0.3% H2O2, for 10 min at 4 C. After incubation with code image was obtained with the support of the flexImaging 2.1
10% fetal calf serum in phosphate buffered saline solution, pri- software (Bruker Daltonik, Germany).
mary antibodies, i.e. mouse anti-CNPase (CNPase, clone SMI91, A supervised analysis was performed on selected spectra from
Sternberger Monoclonals Inc., Lutherville, MD, USA) or mouse anti- regions of interest, predefined in the histopathological images and
PLP (PLP; 1:500, clone: plpc1, Serotec, Oxford, UK), were applied designated as lesion 1 (L1), lesion 2 (L2), white matter (WM) and
overnight. Macrophages/activated microglia were detected by grey matter (GM). The statistical analysis was performed on the ion
KiM1P immunohistochemistry following the protocol described in signal intensities of the average spectrum calculated for each region
Radzun [18]. Antibody binding was detected with biotinylated don- of interest by applying algorithms embedded in the ClinProTools 2.2
key anti-mouse antibodies (Calbiochem), streptavidin-peroxidase software. The distributions of the ion mass signals resulting from
(Sigma) and visualised with diaminobenzidine (Sigma). The sec- the statistical analysis were manually ascertained by displaying the
tions were counterstained with haematoxylin, then dehydrated ion mass distribution within the tissue slides using the flexImag-
and mounted. ing 2.1 software. The ion mass distribution was displayed applying
Thymosin beta-4 was detected with a primary antibody (Rabbit a colour-coded gradient system to highlight the ion mass and its
polyclonal anti-thymosin beta-4-antibody; Peninsula Laborato- abundance.
ries, San Carlos, CA, USA), a biotinylated secondary antibody and
a horseradish-peroxidase/DAB detection system (Super Sensitive 2.8. Protein marker identification
Polymer-HRP IHC Detection System/DAB, BioGenex Fremont, CA,
USA), according to the protocol above. 2.8.1. Sample preparation
White matter and cortical lesion areas of tissue slides (10 m
thickness) from brain block B, indicated in Fig. 3A as L2 and GM*
2.5. Sample preparation for MALDI–IMS
(cortical lesions), respectively, were manually cut. Three L2 and
GM* tissue areas each were pooled and subjected to tissue lysis
The brain tissue was sliced (10 m) using a microtome instru-
◦ and protein extraction protocols. Briefly, the tissue samples were
ment (Leica, CM30050) at −18 C and transferred onto precooled
− ◦ immersed into chilled 0.1% trifluoroacetic acid aqueous solution
( 20 C) Indium-Tin-Oxide (ITO) coated slides (Bruker, Bremen,
(200 l) and ultra-sonicated for 15 min in an ice bath. The homog-
Germany). Afterwards, the slides were warmed on the hand inside ◦
enized sample was centrifuged at 20,000g for 15 min at 4 C. The
the cryostat and then desiccated (45 min) in a vacuum desiccator. ◦
supernatant was collected and stored at −20 C, while the pellet
The dried slides were washed (15 s) in 70% ethanol (Sigma–Aldrich,
was immersed in 50% acetonitrile, 0.1% trifluoroacetic acid aqueous
Germany) and in 96% ethanol (15 s) and then desiccated again
solution (120 l) and ultra-sonicated for 15 min in an ice bath. The
(45 min). The dried tissue slides were wrapped in aluminium foil
◦ extracted proteins were separated by centrifugation at 20,000g for
−
and stored at 80 C. ◦
15 min at 4 C. The supernatant in 50% acetonitrile was lyophilized
◦
and stored at −20 C.
2.6. MALDI–IMS analysis
The L2 and GM* (cortical lesions) extracts in 0.1% trifluoroacetic
acid aqueous solution were washed three times with 50 mM
The tissue slides were scanned with a Nikon Coolscan 5 ED
ammonium acetate using Vivaspin cartridge (CO 3 kDa, Sartorius,
imager (Nikon). The images were used to delimit the measure-
Germany). The samples were centrifuged at 5000 rpm for 30 min
ment area for the MALDI spectra acquisition. Afterwards the tissue
and concentrated down to 20 l.
slides were coated with the matrix using the ImagePrep station
Subsequently, the protein extracts were chromatographed
(Bruker Daltonik, Bremen, Germany) following the manufacturer’s
using a micro column mRP-C18 (4.5 × 50 mm, Agilent Technologies,
◦
protocol. Sinapinic acid (Bruker Daltonik; 10 mg/ml) in 60:40 (v/v)
USA) conditioned at 80 C with an analytical HPLC HP1100 cou-
acetonitrile:water (Merck, Germany) and 0.2% trifluoroacetic acid
pled to an AS1200 fraction collector (Agilent Technologies, USA).
(Fluka, Germany) were used as matrix. Subsequently, a thin gold
For protein separation 0.1% trifluoroacetic acid aqueous solution
film was deposited onto the tissue slides using an Agar Auto Sputter
(solution A) and 0.1% trifluoroacetic acid in acetonitrile (solution
Coater instrument (Plano GmbH, Wetzlar, Germany). MALDI-IMS
B) were used as eluents at a flow rate of 750 l/min using the fol-
TM
measurements were performed using a linear AutoFlex II mass
lowing gradient elution: eluent B, 0–3% in 7 min, 3–22% in 23 min,
spectrometer (Bruker Daltonik, Bremen, Germany) equipped with
22–38% in 5 min, 38–97% in 2 min. The eluate was monitored at
TM
the Smart-beam laser at 100 Hz and laser beam of 85 m diame-
220 nm and 30 s fractions were collected into a 96 microtiter plate.
◦
ter. The spectra were acquired in the linear mode in the mass range
The collected fractions were lyophilized and stored at −20 C.
of 2–30 kDa, lateral resolution of 200–500 m, under control of the
The L2 and GM* (cortical lesions) protein extracts in 50% ace-
flexImaging 2.1 software (Bruker Daltonik). The MALDI instrument
tonitrile were dissolved in Laemmli buffer and separated by sodium
was externally calibrated using the calibration standard mixture
dodecyl sulphate-polyacrylamide gel electrophoresis (15% resolv-
purchased from Bruker Daltonik, Germany.
ing gel and 5% stacking gel) at 120 V. The gel was stained overnight
134 G. Maccarrone et al. / J. Chromatogr. B 1047 (2017) 131–140
Fig. 1. MALDI imaging mass spectrometry (MALDI-IMS) of human tissue samples workflow. The frozen tissue is sectioned and consecutive slides are analysed by (A) histology;
(B) MALDI imaging mass spectrometry (MALDI–IMS). (C) Ion signals (m/z) are plotted with support of flexImage software to generate ion map images with ion mass (m/z),
ion spatial coordinates (x, y) and ion abundance. m/z: mass-to-charge ratio.
with Coomassie Brilliant Blue R250 (Serva, Heidelberg, Germany). 3. Results
After destaining (50% methanol, 12% acetic acid), the gel lanes were
cut in 20 pieces and subjected to in-gel tryptic digestion protocol Histology of brain blocks A and B revealed chronic inactive mul-
as described in Maccarrone [20]. tiple sclerosis lesions with different extents of remyelination and
few foamy macrophages at lesion rims (data not shown). Block
2
A displayed the lesion 1 (51 mm ) constituting of two partially
2
remyelinated areas with 59% remyelination extent (28 mm and 2.8.2. LC–ESI–MSMS
2 2
2 mm ) and lesion 2 (5 mm ) which presented one small area
The protein and peptide identification was performed using a
2
(1 mm ) with 17% remyelination extent (Supplementary Fig. 1A).
nano Eksigent-2D system (Eksigent, Dublin, CA) coupled via a nano
2
Block B presented lesion 1 (5 mm ) almost completely remyeli-
electrospray ion source (Thermo Fisher) to an LTQ-Orbitrap mass
2
nated (96% remyelination extent) and lesion 2 (91 mm ) with
spectrometer (Thermo Fisher Scientific, Bremen, Germany). The
2 2
2 remyelinated areas (27 mm and 5 mm ) equivalent to 35%
samples dissolved in 0.1% formic acid (eluent A) were loaded onto
remyelination extent (Supplementary Fig. 1B). In addition, clus-
an in house-packed nano column (75 m × 15 cm) and eluted by
ters of erythrocytes were detected within lesion 2, indicating small
applying a 95% acetonitrile/0.1% formic acid (eluent B) gradient
blood vessels (Supplementary Fig. 1C).
from 2 to 45% for 40 min at a flow rate of 200 nl/min. The mass
To preserve the tissue morphology, the mass spectrometry anal-
spectrometer was operated in positive mode and data-dependent
ysis and the histological staining were performed on consecutive
scan mode. Full scans were recorded at the Orbitrap mass analyzer
brain tissue sections as depicted in Fig. 1A and B. The overlap of the
at a mass range of m/z 380–1600 and a resolution R = 60,000. The
LFB-stained and unstained images did not show any shifts in the
MS/MS analysis was performed on the five most intense peptide
morphological features (data not shown). The MALDI spectra were
ions and recorded in the LTQ mass analyzer.
first submitted to a multivariate statistical analysis, principal com-
ponent analysis and then to an unsupervised hierarchical clustering
analysis without taking in consideration histo-morphological fea-
2.8.3. Protein identification
tures and molecular targets.
The raw data generated by the mass spectrometer were
The unsupervised analysis of the mass spectra acquired within
searched against SwissProt 15.3 database, taxonomy human, using
brain slide block A generated sub-regions within the tissue area
Mascot search engine (www.matrixscience.com). The data derived
corresponding to normal-appearing white matter, grey matter and
from proteins/peptides extracted by 0.1% trifluoroacetic acid
lesions as depicted in Fig. 2A and Fig. 2C. In block B a phenotypic
were searched applying the following parameters: no enzyme,
composition difference in the area defined as grey matter by LFB-
methionine oxidation, N-terminal acetylation, Ser, Thr and Tyr
histochemistry was detected. In the grey matter area mainly three
phosphorylation as variable modifications. The raw data derived
molecular components have been determined, a group including
from the in-gel tryptic digestion samples were searched applying
two molecular components, cornflower-blue and yellow, consti-
the following parameters: trypsin as enzyme, 2 missed cleav-
tutes one of the grey matter areas and another group including the
ages, methionine oxidation, Ser, Thr and Tyr phosphorylation and
molecular components yellow and brown mainly defines the other
variable modifications, cystein carboxymethylation as static mod-
grey matter area (Fig. 2D,F). The immunostainings for CNPase and
ification. For protein database searches the mass accuracies for
PLP performed on the block B detected widespread subpial cortical
precursor and fragment ions were set to 20 ppm and 0.6 Da, respec- tively.
G. Maccarrone et al. / J. Chromatogr. B 1047 (2017) 131–140 135
Fig. 2. Optical images of LFB-stained brain tissue slides. Hierarchical clustering dendrograms and respective images generated by MALDI imaging mass spectrometry
(MALDI–IMS) data (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).
Block A, brain slide: (A) LFB-stained image, lesion 1 (L1), lesion 2 (L2), white matter (WM) and grey matter (GM) regions are indicated. (B) Dendrogram of MALDI spectra
acquired across block B slide. (C) The reconstruction of the top five dendrogram branches reflects white matter (green and violet), lesion 1 (sky-blue), grey matter (brown)
and a sub-area (pink) within lesion 2 (green).
Block B, brain slide: (D) LFB-stained image, lesion 1 (L1), lesion 2 (L2), white matter (WM) and grey matter (GM) regions are indicated. (E) Dendrogram of MALDI spectra
acquired across block A slide. (F) MALDI-IMS image is generated from the reconstruction of the top nodes of the dendrogram branches. Two sub-areas within lesion 2 (violet
and green), two components within lesion 1 (cornflower-blue and pink) are identified. The white matter (WM) mainly consists of one component (pink). The grey matter
area is constituted of three molecular components, a group of two components (cornflower-blue and yellow) defined one area, another group of two components (yellow
and brown) mainly identified the other grey matter area.
The construction of the dendrogram branches and the corresponding ion mass images are displayed using a colour-coded system. The image pixels are coloured according
to the colour of the respective dendrogram node.
Fig. 3. LFB-staining, PLP and CNPase immunohistochemical images. Optical images of block B brain slide, (A) LFB-staining, (B) PLP, (C) CNPase. Cortical lesions (CL, GM*) and
white matter lesions (L), normal appearing white matter (WM) and grey matter (GM) are indicated.
demyelination (cortical lesions) in the grey matter area. The cortical Fig. 2. The molecular masses m/z 15205 and 15940 are specifi-
lesions are mainly located in the areas labelled GM* (Fig. 3A–C). cally distributed within the BV area (Fig. 4A). We hypothesized
To determine the molecular composition within the brain the presence of human haemoglobin isoforms. This finding was
lesions and the sub-areas detected by the unsupervised analy- confirmed and validated by hematoxylin and eosin staining which
sis, regions of interest based on the tissue histo-morphology were revealed erythrocyte clusters in the BV area indicating the pres-
defined. In block A lesion 1, lesion 2, BV and ROI3, a region at the ence of blood vessels (Supplementary Fig. 1C). The mass m/z 18,608,
rim of lesion 1, areas were defined (Fig. 4A). The analysis revealed which marked the white matter area and part of lesion 2 (Fig. 4A),
molecular masses smaller than 6 kDa in lesion 1 area (17% remyeli- corresponds to isoform 5 of Myelin Basic Protein (18591 Da).
nation), while within lesion 2 (59% remyelination) compounds with In block slide B, lesion 1, lesion 2, normal-appearing white mat-
molecular masses greater than 10 kDa were detected (Table 2). ter, grey matter and GM* (cortical lesions) areas were defined
The ion density maps are depicted in Fig. 4A and Supplementary (Fig. 4B). The supervised statistical analysis of MALDI-IMS spectra
136 G. Maccarrone et al. / J. Chromatogr. B 1047 (2017) 131–140
Fig. 4. Images of ion density maps generated by MALDI imaging mass spectrometry (MALDI–IMS). (A) LFB-stained block A slide lesion 1 (L1), lesion 2 (L2), white matter
(WM) and grey matter (GM) are indicated. The BV area is delimited with black dotted lines. MALDI-IMS ion density maps of mass signals (m/z) expressed within the WM,
BV, L1 and L2 areas are shown. (B) LFB-stained block B, white matter (WM), grey matter (GM), lesion 1 (L1), lesion 2 (L2) and GM* (CL: cortical lesions) areas are indicated.
MALDI–IMS ion density maps of mass signals (m/z) expressed within WM, GM* (CL), L1 and L2 areas are shown.
The image resolution was set to 200 m. The ion abundances are shown based on a colour-coded gradient system.
acquired across lesion 1, lesion 2 and GM* (cortical lesions) areas tary Fig. 4. To identify the molecular masses, the protein extracts
resulted in a list of compounds whose mass values are listed in from the block B lesion 2 and cortical lesions (GM*) were sub-
Table 2. Lesion 1 (96% remyelination) and the normal-appearing jected to proteomics workflow based on LC–ESI–MSMS analysis.
white matter regions are defined mostly by high molecular mass The mass m/z 5000, mapped by MALDI–IMS within block B lesion
compounds m/z 16,473 and m/z 18,650 as depicted by the respec- 2 core and rim (Fig. 4B), was identified as thymosin beta-4 (m/z
tive ion density maps shown in Fig. 4B and Supplementary Fig. 3. 5000, phosphorylated form) and confirmed by anti-thymosin beta-
The lesion 2 area (35% remyelination) is described by two groups 4 immunohistochemical staining (Fig. 5A). In block A the mass m/z
of molecular compounds, demyelinated core of lesion 2 (m/z 3547, 5000 was detected by MALDI-IMS mainly in lesion 1 and partially
3779, 3805, 3822, 4054, 4873, 5000, 5103, 5210) and lesion 2 rim in grey matter (Fig. 4A). The immunohistochemistry analysis con-
(m/z 3340, 5000, 5233, 17364, 17574). The ion density maps of firmed the presence of thymosin beta-4 in block A lesion 1 and
the two groups are shown in Fig. 4B and Supplementary Fig. 3. In in grey matter areas (Fig. 5B a). To validate the identification of
the GM* (cortical lesions) low molecular weight compounds (m/z thymosin beta-4, 12 brain sections from distinct multiple sclerosis
2225, 2569, 2622, 2964, 3340, 3805, 7524, 8475) were found; the patients were analysed by MALDI-IMS. The thymosin beta-4 mass
respective ion density maps are depicted in Fig. 4B and Supplemen- was detected in 12 patients. For validation of anti-thymosin beta-4
G. Maccarrone et al. / J. Chromatogr. B 1047 (2017) 131–140 137
Fig. 5. Thymosin beta-4 (Tb4) identification by MALDI Imaging mass spectrometry (MALDI–IMS) and validation by immunohistochemical analysis. (A) Identification. The
workflow includes thymosin beta-4 mass detection by MALDI–IMS, identification by LC–ESI–MSMS proteomics analysis and assessment by immunohistochemistry. (B)
Validation. MALDI–IMS ion maps, LFB-stained and anti-thymosin beta-4 immuno histochemical images of six out of eight distinct brain slides are reported. One sample
group includes demyelinated and/or active lesions, the second group includes remyelinated lesions. The results of the histopathologic analysis of the brain slides are: (a) Two
demyelinated/only partially remyelinated lesions; (b) Slightly hypocellular, at the border hypercellular lesion with myelin preservation; (c) Inflamed vessel in the centre of
the slight; hypercellular lesions with myelin preservation; (d) Black line: remyelinated lesion; red line: very small demyelination; (e) Completely remyelinated lesion; (f)
Completely remyelinated lesion. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Fig. 6. Immunohistochemical stained human brain tissue slide of multiple sclerosis patient. Thymosin beta-4 (brown dots) is mostly expressed in (A) neurons and (B)
microglia/macrophages (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).
138 G. Maccarrone et al. / J. Chromatogr. B 1047 (2017) 131–140
Table 2
Mass values (m/z) of ion species identified by MALDI imaging mass spectrometry in the multiple sclerosis human brain lesions.
Block A Block A Block A Block A Block B Block B Block B
Lesion 1 (17%) Lesion 2 (59%) BV ROI3 Lesion 1 (96%) Lesion 2 (35%) GM*
(Cortical lesion)
3549 10680 15205 2230 2225 3340 2225
3780 10886 15940 2572 2424 3547 2424
3823 14084 2626 2622 3779 2569
4054 16445 2967 16474 3805 2622
4092 17345 18650 3822 2964
5000 17542 4054 3340
5212 18608 4873 3805
5231 18813 5001 7524
5103 8475
5210 17364
5233 17574 17364
17574
Ion masses generated by the supervised statistical analysis of MALDI-IMS data obtained from the analysis of block A and B brain slides. The value of remyelination extents
are reported in percentage.
by immunohistochemistry serial brain slides from eight patients which were not detectable by routine histology, but were con-
were used (Fig. 5B). Thymosin beta-4 was detected by immuno- firmed by immunostaining with antibodies against CNPase and PLP
histochemistry at the rim of demyelinated and/or chronic active (Fig. 3B,C). These results highlight the potential of the unsupervised
lesions, in grey matter and highly abundant around an inflamed approach for hypothesis-free MALDI-IMS data analysis to identify
blood vessel in one of the brain slides (Fig. 5B). Thymosin beta-4 abnormalities that are not detectable by routine histopathology.
is not present in fully remyelinated lesion as depicted in Fig. 5B The concordance of MALDI-IMS ion mass images with histopathol-
(remyelinated lesions). In-depth analysis of lesions revealed that ogy images was demonstrated by the discovery of the sub-area BV
thymosin beta-4 is mainly expressed in macrophages and activated within the block A slide, mapped by the compounds, m/z 15,205
microglia (Fig. 6). In the normal-appearing tissue, thymosin beta-4 and m/z 15,940 characteristic for haemoglobin. Indeed, the hema-
was mainly detected in the grey matter, but not in oligodendro- toxylin and eosin stain revealed in the discrete area BV the presence
cytes. of erythrocytes in the tissue (Supplementary Fig. 1C).
The goal of our study was to determine proteins and peptides
4. Discussion specifically expressed in the normal-appearing white matter, grey
matter and in the lesions. To achieve this aim we interrogated the
In this study, for the first time the MALDI imaging technology is MALDI imaging spectra of white matter lesions, cortical lesions and
applied to map proteins and peptides expressed in human multiple normal-appearing white matter area and compared those using
sclerosis brain lesions. The determination of molecular masses and statistical tools.
spatial coordinates of proteins specifically expressed in the lesions The statistical analysis resulted in protein and peptide mass sig-
leads to the identification and localization of potential biomarkers nals whose pattern profiles distinguished the areas with different
involved in the tissue degradation or regeneration. These find- myelin content. Indeed, the molecular compositions were specifi-
ings can advance the knowledge in understanding mechanisms cally correlated with the histological features and allowed unbiased
involved lesion formation or regeneration. discrimination between the different partially remyelinated lesions
Recently, proteomics studies performed on active multiple scle- in both brain slides. The large overlap between the molecular com-
rosis lesions revealed targets for new therapies [14]. However, pounds present in the lesions of the two brain slides validated the
since the proteomics approach was based on the disruption of the link between analytes and tissue anatomical features. The lesions
tissue, these data lacked the proteins’ spatial distribution which with low extents of remyelination showed a marked overlap of
is important information for assessing the correlation between seven low molecular weight compounds (Table 2). As expected,
anatomical features, progress and development of the disease. The insufficiently remyelinated plaques were composed of low molec-
MALDI-IMS approach used in the present study enables the simul- ular weight components, possibly generated from degradation
taneous detection of molecular masses, their spatial distributions processes or caused by a lack of myelin proteins. In block B lesion
and abundances derived from multiple analytes without destroy- 1 with 96% remyelination extent, low molecular weight specimens
ing the tissue morphology. The method is unbiased and allows to were detected in addition to high molecular mass components. We
discover and to evaluate in parallel multiple compounds within the hypothesize that they are derived from protein breakdown occur-
same tissue section. This is in contrast to immunohistochemistry ring during the lesion formation. The block A lesion 2 with 59%
which requires target-specific reagents and multiple tissue sam- remyelination consisted of components with molecular weights
ples. The MALDI-IMS data nicely complement results obtained with of 10–18 kDa, which indicates that the remyelination process was
immunohistochemistry. incomplete or in progress. Interestingly, the protein profiles of cor-
In this study we focused on evaluating the ability of MALDI- tical lesions showed a similar pattern of compounds as the white
IMS to determine multiple analytes linked to brain tissue areas matter lesions with low myelin content (Table 2). Although corti-
with different myelin contents. Initially, we performed a hierarchi- cal lesions have been previously shown to differ from white matter
cal clustering analysis and classification of the MALDI-IMS dataset lesions by lower inflammation and mild astrogliosis, the MALDI-
without prior knowledge of the histological image. The unsuper- IMS results suggest that similar degradation processes may occur
vised analysis resulted in classes of compounds which presented within multiple sclerosis lesions of the white matter and grey mat-
a large correlation to white matter, grey matter and lesion areas. ter [21,22]. These findings are consistent with the hypothesis that
Interestingly, the in-depth interrogation of the unsupervised hier- cortical lesions are caused by similar autoimmune processes as
archical clustered MALDI dataset led us to discover cortical lesions white matter lesions [23].
G. Maccarrone et al. / J. Chromatogr. B 1047 (2017) 131–140 139
Another objective of this study was to identify peptides and multiple sclerosis lesions is demonstrated. Our data suggest a neu-
proteins expressed in the white matter and cortical lesions. For roprotective and neurorestorative role of thymosin beta-4 in the
this purpose we applied a proteomics workflow including liquid remodeling and remyelination processes occurring in the central
chromatography and electrospray ionisation tandem mass spec- nervous system of multiple sclerosis patients. However, to explore
trometry [24,25]. The proteomic analysis led to the identification the underlying mechanisms of remyelination in multiple sclerosis
of thymosin beta-4 whose mass (m/z 5000) was previously detected further studies on the identification of the molecular compounds
by the supervised statistical analysis within white matter partially in-situ are warranted.
remyelinated lesions (Fig. 4A,B). Immunohistochemistry of eight
multiple sclerosis cohort brain sections assessed that thymosin Funding
beta-4 is mainly expressed in macrophages and activated microglia
located at the rim of chronic active and/or demyelinated lesions This work was supported by the Max Planck Society (www.
(and in the grey matter) (Figs. 5 B, 6). In remyelinated lesions, mpg.de). This work was partly founded by Novartis Pharma GmbH,
thymosin beta-4 is detected neither by MALDI-IMS nor by immuno- Germany.
histochemistry as depicted in Fig. 5B (remyelinated lesions).
Thymosin beta-4 is a peptide that plays an important role in Acknowledgments
cell migration, in recruitment of different progenitor cells and in
the recovery processes in different tissues. Its main function is the The authors acknowledge the UK Multiple Sclerosis Tissue Bank,
promotion of cell migration by chemotactic effects and modula- Division of Neuroscience and Mental Health Imperial College, Lon-
tion of the cytoskeleton by sequestration of G actin monomers don for providing the brain tissue, Dr. Michael Becker (Bruker
[26]. As thymosin beta-4 is expressed in different immune cells, Daltonik GmbH) for supporting the imaging data analysis and Clau-
a role of the peptide in immune reactions has been postulated dia Kühne (Max-Planck Institute of Psychiatry) for recording the
[27]. During oxidative stress thymosin beta-4 has been shown to close-up images of the brain lesions and the assistent student
exhibit inhibitory effects on inflammation [26]. In our study thy- Korbinian Ehrmann (Max-Planck Insitute of Psychiatry) for sup-
mosin beta-4 is expressed in the grey matter. This observation porting in sample preparation for protein identification.
is similar to the results reported on human central nervous sys-
tem tissue not affected by multiple sclerosis [28]. The presence
Appendix A. Supplementary data
of thymosin beta-4 at the rim of not yet completely remyelinated
multiple sclerosis lesions may hint to a thymosin beta-4 role in
Supplementary data associated with this article can be found,
restorative effects in the neighborhood of normal appearing white
in the online version, at http://dx.doi.org/10.1016/j.jchromb.2016.
matter. In multiple sclerosis mouse model it was observed that thy- 07.001.
mosin beta-4 treatment have a protective role for the inhibition of
inflammation during the acute process and later in the recruitment References
of oligodendrocyte progenitor cells to support the remyelination
processes [29,30]. Thymosin beta-4 has been shown to be reduced [1] A. Compston, A. Coles, Multiple sclerosis, Lancet 372 (2008) 1502–1517.
[2] F. Barkhof, W. Bruck, C.J. De Groot, E. Bergers, S. Hulshof, J. Geurts, C.H.
in the cerebrospinal fluid of patients with clinical multiple scle-
Polman, P. van der Valk, Remyelinated lesions in multiple sclerosis: magnetic
rosis compared to those with a clinical isolated syndrome, which
resonance image appearance, Arch. Neurol. 60 (2003) 1073–1081.
indicates anti-inflammatory and regenerative effects in very early [3] R. Patani, M. Balaratnam, A. Vora, R. Reynolds, Remyelination can be extensive
in multiple sclerosis despite a long disease course, Neuropathol. Appl.
disease stages [31]. In stroke an upregulation of thymosin beta-
Neurobiol. 33 (2007) 277–287.
4 RNA has been shown in macrophages, but also in hippocampal
[4] P. Patrikios, C. Stadelmann, A. Kutzelnigg, H. Rauschka, M. Schmidbauer, H.
neurons, supporting neurite extension in surviving neurons [32]. Laursen, P.S. Sorensen, W. Brück, C. Lucchinetti, H. Lassmann, Remyelination
is extensive in a subset of multiple sclerosis patients, Brain 129 (2006)
A neuro-restorative effect of thymosin beta-4 treatment has been
3165–3172.
shown in a rat model of neuronal injury [33,34], in a rat model
[5] M.R. Kotter, C. Stadelmann, H.P. Hartung, Enhancing remyelination in
of stroke [35] and in an experimental allergic encephalomyelitis disease—can we wrap it up? Brain 134 (2011) 1882–1900.
mouse model for multiple sclerosis [30]. Thymosin beta-4 treat- [6] W. Brück, N. Sommermeier, M. Bergmann, U. Zettl, H.H. Goebel, H.A.
Kretzschmar, Macrophages in multiple sclerosis, Immunobiology 195
ment in these animal models showed a reduction of inflammation.
(October 4–5) (1996) 588–600.
Improved remyelination is revealed by an increased recruitment
[7] V.E. Miron, R.J. Franklin, Macrophages and CNS remyelination, J. Neurochem.
of oligodendrocyte progenitor cells and a high number of mature 130 (2014) 165–171.
[8] M. Besser, R. Wank, Cutting edge: clonally restricted production of the
oligodendrocytes within the injured tissue which led to a better
neurotrophins brain-derived neurotrophic factor and neurotrophin-3 mRNA
functional outcome [33,30]. Thymosin beta-4 is involved in the
by human immune cells and Th1/Th2-polarized expression of their receptors,
remyelination process by upregulating p38 mitogen-activated pro- J. Immunol. 162 (1999) 6303–6306.
[9] M. Torcia, L. Bracci-Laudiero, M. Lucibello, L. Nencioni, D. Labardi, A.
tein kinase which controls oligodendrocyte differentiation leading
Rubartelli, F. Cozzolino, L. Aloe, E. Garaci, Nerve growth factor is an autocrine
to oligodendrocyte maturation and myelin gene expression [36].
survival factor for memory B lymphocytes, Cell 85 (1996) 345–356.
[10] M. Kerschensteiner, E. Gallmeier, L. Behrens, V.V. Leal, T. Misgeld, W.E.
Klinkert, R. Kolbeck, E. Hoppe, R.L. Oropeza-Wekerle, I. Bartke, C. Stadelmann,
H. Lassmann, H. Wekerle, R. Hohlfeld, Activated human T cells, B cells, and
5. Conclusions
monocytes produce brain-derived neurotrophic factor in vitro and in
inflammatory brain lesions: a neuroprotective role of inflammation? J. Exp.
Med. 189 (1999) 865–870.
In conclusion we could demonstrate that MALDI-IMS has the
[11] R. Hohlfeld, M. Kerschensteiner, C. Stadelmann, H. Lassmann, H. Wekerle, The
ability to distinguish lesions with different extents of remyeli-
neuroprotective effect of inflammation: implications for the therapy of
nation, to confirm the lesion “classification” determined by the multiple sclerosis, Neurol. Sci. 27 (2006) S1–7.
histopathology and to look beyond histology. It allowed the [12] J.B. Leverenz, I. Umar, Q. Wang, T.J. Montine, P.J. McMillan, D.W. Tsuang, J. Jin,
C. Pan, J. Shin, D. Zhu, J. Zhang, Proteomic identification of novel proteins in
discovery of cortical lesions which were not visible in routine his-
cortical lewy bodies, Brain Pathol. 17 (2007) 139–145.
tochemistry. This study highlights the major strengths and the
[13] A.M. Jastorff, K. Haegler, G. Maccarrone, F. Holsboer, F. Weber, T. Ziemssen,
abilities of MALDI-IMS in performing an unbiased molecular map- C.W. Turck, Regulation of proteins mediating neurodegeneration in
experimental autoimmune encephalomyelitis and multiple sclerosis,
ping of multiple sclerosis lesions and enables the identification
Proteomics Clin. Appl. 3 (2009) 1273–1287.
of a potential disease relevant protein, thymosin beta-4. For the
[14] M.H. Han, S.I. Hwang, D.B. Roy, D.H. Lundgren, J.V. Price, S.S. Ousman, G.H.
first time the presence of endogenous thymosin beta-4 in human Fernald, B. Gerlitz, W.H. Robinson, S.E. Baranzini, B.W. Grinnell, C.S. Raine, R.A.
140 G. Maccarrone et al. / J. Chromatogr. B 1047 (2017) 131–140
Sobel, D.K. Han, L. Steinman, Proteomic analysis of active multiple sclerosis [25] S. Rauser, C. Marquardt, B. Balluff, S.O. Deininger, C. Albers, E. Belau, R.
lesions reveals therapeutic targets, Nature 451 (2008) 1076–1081. Hartmer, D. Suckau, K. Specht, M.P. Ebert, M. Schmitt, M. Aubele, H. Höfler, A.
[15] M.D. Johnson, J.L. Floyd, R.M. Caprioli, Proteomics in diagnostic Walch, A classification of HER2 receptor status in breast cancer tissues by
neuropathology, J. Neuropathol. Exp. Neuro. 165 (2006) 837–845. MALDI imaging mass spectrometry, J. Proteome Res. 9 (2010) 1854–1863.
[16] J. Franck, K. Arafah, M. Elayed, D. Bonnel, D. Vergara, A. Jacquet, D. Vinatier, M. [26] T. Huff, C.S. Müller, A.M. Otto, R. Netzker, E. Hannappel, beta-Thymosins,
Wisztorski, R. Day, I. Fournier, M. Salzet, MALDI imaging mass spectrometry: small acidic peptides with multiple functions, Int. J. Biochem. Cell Biol. 33
state of the art technology in clinical proteomics, Mol. Cell. Proteomics 8 (2001) 205–220.
(2009) 2023–2033. [27] H. Gondo, J. Kudo, J.W. White, C. Barr, P. Selvanayagam, G.F. Saunders,
[17] J. Pierson, J.L. Norris, H.R. Aerni, P. Svenningsson, R.M. Caprioli, P.E. Andrén, Differential expression of the human thymosin-beta 4 gene in lymphocytes
Molecular profiling of experimental Parkinson’s disease: direct analysis of macrophages, and granulocytes, J. Immunol. 139 (1987) 3840–3848.
peptides and proteins on brain tissue sections by MALDI mass spectrometry, J. [28] J. Hanrieder, T. Ekegren, M. Andersson, J. Bergquist, MALDI imaging of
Proteome Res. 3 (2004) 289–295. post-mortem human spinal cord in amyotrophic lateral sclerosis, J.
[18] H.J. Radzun, M.L. Hansmann, H.J. Heidebrecht, S. Bödewadt-Radzun, H.H. Neurochem. 124 (2013) 695–707.
Wacker, H. Kreipe, H. Lumbeck, C. Hernandez, C. Kuhn, M.R. Parwaresch, [29] R. Reti, E. Kwon, P. Qiu, M. Wheater, G. Sosne, Thymosin beta4 is
Detection of a monocyte/macrophage differentiation antigen in routinely cytoprotective in human gingival fibroblasts, Eur. J. Oral Sci. 116 (2008)
processed paraffin-embedded tissues by monoclonal antibody Ki-M1P, Lab. 424–430.
Invest. 65 (1991) 306–315. [30] J. Zhang, Z.G. Zhang, D. Morris, Y. Li, C. Roberts, S.B. Elias, M. Chopp,
[19] S.O. Deininger, M.P. Ebert, A. Fütterer, M. Gerhard, C. Röcken, MALDI imaging Neurological functional recovery after thymosin beta4 treatment in mice with
combined with hierarchical clustering as a new tool for the interpretation of experimental auto encephalomyelitis, Neuroscience 164 (2009) 1887–1893.
complex human cancers, J. Proteome Res. 7 (2008) 5230–5236. [31] M. Liguori, A. Qualtieri, C. Tortorella, V. Direnzo, A. Bagala, M. Mastrapasqua,
[20] G. Maccarrone, I. Birg, E. Malisch, M.C. Rosenhagen, C. Ditzen, J.A. Chakel, F. P. Spadafora, M. Trojano, Proteomic profiling in multiple sclerosis clinical
Mandel, C.-C. Doertbudak, K. Haegler, F. Holsboer, C.W. Turck, In-depth courses reveals potential biomarkers of neurodegeneration, PLoS One (2014)
analysis of the human CSF proteome using protein prefractionation, Clin. e103984.
Proteomics 1 (2004) 333–364. [32] N. Vartiainen, I. Pyykönen, T. Hökfelt, J. Koistinaho, Induction of thymosin
[21] C. Stadelmann, C. Wegner, W. Brück, Inflammation demyelination, and beta(4) mRNA following focal brain ischemia, Neuroreport 7 (1996)
degeneration—recent insights from MS pathology, Biochim. Biophys. Acta 1613–1616.
1812 (2011) 275–282. [33] D.C. Morris, Z.G. Zhang, J. Zhang, Y. Xiong, L. Zhang, M. Chopp, Treatment of
[22] B.P. Brink, R. Veerhuis, E.C. Breij, P. van der Valk, C.D. Dijkstra, L. Bö, The neurological injury with thymosin 4, Ann. N. Y. Acad. Sci. 1269 (2012)
pathology of multiple sclerosis is location-dependent: no significant 110–116.
complement activation is detected in purely cortical lesions, J. Neuropathol. [34] Y. Xiong, Y. Zhang, A. Mahmood, Y. Meng, Z.G. Zhang, D.C. Morris, M. Chopp,
Exp. Neurol. 64 (2005) 147–155. Neuroprotective and neurorestorative effects of thymosin 4 treatment
[23] A. Kutzelnigg, C.F. Lucchinetti, C. Stadelmann, W. Brück, H. Rauschka, M. initiated 6 hours after traumatic brain injury in rats, J. Neurosurg. 116 (2012)
Bergmann, M. Schmidbauer, J.E. Parisi, H. Lassmann, Cortical demyelination 081–092.
and diffuse white matter injury in multiple sclerosis, Brain 128 (2005) [35] D.C. Morris, M. Chopp, L. Zhang, M. Lu, Z.G. Zhang, Thymosin beta4 improves
2705–2712. functional neurological outcome in a rat model of embolic stroke,
[24] M.R. Groseclose, M. Andersson, W.M. Hardesty, R.M. Caprioli, Identification of Neuroscience 169 (2010) 674–682.
proteins directly from tissue: in situ tryptic digestions coupled with imaging [36] M. Santra, M. Chopp, Z.G. Zhang, M. Lu, S. Santra, A. Nalani, S. Santra, D.C.
mass spectrometry, J. Mass Spectrom. 42 (2007) 254–262. Morris, Thymosin  4 mediates oligodendrocyte differentiation by
upregulating p38 MAPK, Glia 60 (2012) 1826–1838.