Page 1
RESEARCH Open Access
Quantitative automated microscopy (QuAM)
elucidates growth factor specific signalling in
pain sensitization
Christine Andres1,2, Sonja Meyer3, Olayinka A Dina4, Jon D Levine4, Tim Hucho1*
Abstract
Background: Dorsal root ganglia (DRG)-neurons are commonly characterized immunocytochemically. Cells are
mostly grouped by the experimenter’s eye as “marker-positive” and “marker-negative” according to their
immunofluorescence intensity. Classification criteria remain largely undefined. Overcoming this shortfall, we
established a quantitative automated microscopy (QuAM) for a defined and multiparametric analysis of adherent
heterogeneous primary neurons on a single cell base.
The growth factors NGF, GDNF and EGF activate the MAP-kinase Erk1/2 via receptor tyrosine kinase signalling. NGF
and GDNF are established factors in regeneration and sensitization of nociceptive neurons. If also the tissue
regenerating growth factor, EGF, influences nociceptors is so far unknown. We asked, if EGF can act on nociceptors,
and if QuAM can elucidate differences between NGF, GDNF and EGF induced Erk1/2 activation kinetics. Finally, we
evaluated, if the investigation of one signalling component allows prediction of the behavioral response to a
reagent not tested on nociceptors such as EGF.
Results: We established a software-based neuron identification, described quantitatively DRG-neuron heterogeneity
and correlated measured sample sizes and corresponding assay sensitivity. Analysing more than 70,000 individual
neurons we defined neuronal subgroups based on differential Erk1/2 activation status in sensory neurons. Baseline
activity levels varied strongly already in untreated neurons. NGF and GDNF subgroup responsiveness correlated
with their subgroup specificity on IB4(+)- and IB4(-)-neurons, respectively. We confirmed expression of EGF-
receptors in all sensory neurons. EGF treatment induced STAT3 translocation into the nucleus. Nevertheless, we
could not detect any EGF induced Erk1/2 phosphorylation. Accordingly, intradermal injection of EGF resulted in a
fundamentally different outcome than NGF/GDNF. EGF did not induce mechanical hyperalgesia, but blocked PGE2-
induced sensitization.
Conclusions: QuAM is a suitable if not necessary tool to analyze activation of endogenous signalling in
heterogeneous cultures. NGF, GDNF and EGF stimulation of DRG-neurons shows differential Erk1/2 activation
responses and a corresponding differential behavioral phenotype. Thus, in addition to expression-markers also
signalling-activity can be taken for functional subgroup differentiation and as predictor of behavioral outcome. The
anti-nociceptive function of EGF is an intriguing result in the context of tissue damage but also for understanding
pain resulting from EGF-receptor block during cancer therapy.
Background
A common denominator of DRG-neurons is their extreme
heterogeneity. They differ in respect to parameters such as
morphology, protein expression and functionality [1]. To
what extent functional differences can be correlated to
expression differences of e.g. ion channels is a matter of
intense research. Mostly the expression of a “marker” is
detected via immunofluorescence microscopy [2]. The
grouping into “marker-positive” and “marker-negative”
cells is commonly performed qualitatively by eye by a
trained experimenter. That a differentiation into a “posi-
tive” and “negative” population can be accomplished by
* Correspondence: hucho@molgen.mpg.de
1Department for Molecular Human Genetics, Max Planck Institute for
Molecular Genetics, Ihnestrasse 73, 14195 Berlin, Germany
Full list of author information is available at the end of the article
Andres et al. Molecular Pain 2010, 6:98
http://www.molecularpain.com/content/6/1/98
MOLECULAR PAIN
© 2010 Andres et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.
Page 2
eye is mostly only assumed but not experimentally
addressed. Even further, evaluation through an experimen-
ter’s eye does not allow a definition of exact quantitative
parameters such as which staining intensity qualifies as
“positive”. Thereby comparability between different labs,
experimenters and/or experimental days cannot be ana-
lyzed and has to be questioned. The inherent inaccuracy
of the “by eye” evaluation of intensities might underlie the
wide variety of population sizes reported in the literature.
For example the population size of IB4(+)-neurons ranges
from 40 - 70% [3-5] and population size for TrkA(+)-neu-
rons ranges from 35 - 70% [3,6]). In addition, more com-
plex evaluations such as the gradual increase in activation
of signalling components are near to impossible. This is in
times of highly sophisticated means of quantification
increasingly unsatisfying and problematic. Thus, we set
out to establish a technique to detect and quantify immu-
nofluorescence signals of adherent primary sensory neu-
rons on a single cell base.
Functionality of sensory neurons is mostly addressed
by investigation of ion channel properties and differen-
tial ion channel expression. But the sensitivity of ion
channels can be modulated to a great extend by post
translational modification such as phosphorylation in
the course of intracellular signalling cascade events
[1,7-9]. Therefore activation of a receptor not necessarily
results in the activation of its downstream signalling.
Thus the expression of a receptor cannot be taken as
synonymous with functional changes in these neurons.
To what extent the activation properties of intracellular
signalling cascades themselves can be taken to differenti-
ate functional different neuronal subgroups has not been
addressed so far. Thus we tested if we can detect gra-
dual increase in signalling component activation in sub-
groups of sensory neurons and if investigation of such
cellular signalling kinetics allows to predict if a sub-
stance has a sensitizing effect in behavioral experiments.
Growth factors play a dual role in sensory neurons. In
tissue challenged with potentially damaging stimuli they
mediate neuronal tissue defense and alarm of the organ-
ism. NGF is a potent endogenous stimulator of neuronal
survival and nerve fiber growth [10-12] and is essential
for reinnervation of the skin after injury to cutaneous
nerves [13,14]. But NGF also initiates and maintains
hypersensitivity [15]. Also the growth factor, GDNF,
shows this dual effect. On one hand GDNF is an essen-
tial growth factor for the survival and functional mainte-
nance of a subgroup of nociceptors during development
and after insult [16,17]. On the other hand GDNF has
been reported to contribute to inflammatory hyperalge-
sia in an adjuvant-induced pain model [18] and also
acutely if injected into the skin [19].
On the cellular level, growth factors bind and activate
receptor tyrosine kinase receptors, resulting among
others in the activation of the MAPK Erk1/2 [20,21].
Accordingly, GDNF and NGF-induced mechanical pain
sensitization has been identified to depend on the acti-
vation of Erk1/2 [19,22].
Beyond the known effects of the growth factors NGF
and GDNF on nociceptors, there is another family of
growth factors central to tissue protection and wound
healing, the EGF-family [23]. They are secreted after
tissue injury by platelets, macrophages and fibroblasts
to initiate proliferation and regeneration of the epithe-
lium [24,25]. Wound healing can be improved by EGF
administration [26,27]. Like NGF, EGF exerts its cellu-
lar action via a member of the receptor tyrosine kinase
family, the EGFR. The receptors of NGF and EGF
share many structural features as well as activation
associated intracellular signalling cascades. One signal-
ling component reported in non-nociceptive cells to be
central to EGF-as well as NGF/GDNF signalling is
again Erk1/2 [28].
While described as a survival and growth factor, it is
not known, if EGF has the potential to act on nocicep-
tive neurons. Accordingly, it is also unknown if EGF like
NGF and GDNF sensitizes them. As all three growth
factors act by activation of Erk1/2 this appears likely.
But of special interest is a study in PC12 cells indicating
that the outcome of NGF and EGF treatment can be
very different. While stimulation with NGF leads to
neurite outgrowth and differentiation, stimulation with
EGF leads to the diametral opposing cellular response,
namely cell proliferation [29]. Interestingly, both actions
were described to be mediated by activation of Erk1/2.
How the very same signalling component can result in
two opposing phenotypes has been enigmatic for long.
Only recently the opposing phenotypic outcome of NGF
and EGF signalling was correlated to a differential sti-
mulation kinetic of Erk1/2 phosphorylation induced by
NGF and EGF. While NGF results in a slower but more
pronounced and long-lasting activation, EGF activates
Erk1/2 earlier but only transiently [30]. Thus a cellular
phenotypic outcome can only in part be derived from
the activation of mediating signalling components as
such. Equally important aspects are the kinetic para-
meters like the amplitude, the pace of activation-onset
as well as the duration of activation. Due to the lack of
appropiate techniques it has not been analyzed in noci-
ceptive neurons if there are differences in the kinetics of
Erk1/2 in response to NGF versus EGF.
Therefore‚ we investigated if nociceptive neurons
express the EGF-receptor and if EGF results in activation
of Erk1/2 similar to the growth factors NGF and GDNF.
We thereby tested the concept if the comparison of cellu-
lar activation responses can be used for predicting if EGF
sensitizes nociceptors in behavioral experiments. For
investigating signalling kinetics of endogenous signalling
Andres et al. Molecular Pain 2010, 6:98
http://www.molecularpain.com/content/6/1/98
Page 2 of 16
Page 3
components of nociceptive neurons on single neuron
base we introduced a quantitative automated microscopy
(QuAM).
Methods
Chemicals
BSA, L-glutamine, poly L-ornithine hydrochloride,
DMSO, paraformaldehyde, Triton X-100 and glutamate
were purchased from Sigma (Taufkirchen, Germany),
collagenase P from Roche (Mannheim, Germany), tryp-
sin from Worthington Biochemical Corporation (Free-
hold, NJ, USA), Neurobasal A (without phenol red), B27
supplement, laminin, minimum essential medium with
glutamax were purchased from Invitrogen (Germany,
UK), DMEM, trypsin and EDTA from Clonetics (Cam-
brex, US) and normal donkey serum from Dianova
(Hamburg, Germany).
Drugs
PMA, EGF, PGE2 and epinephrine were purchased from
Sigma (Taufkirchen, Germany or St. Louis, MO). mNGF
was purchased from Alomone (Jerusalem, Israel), GDNF
was purchased from PeproTech (Hamburg, Germany).
Antibodies
Anti-PGP 9.5 was purchased from MorphoSys AG
(Martinsried/Planegg, Germany).
Anti-phospho-Erk (Thr-202/Tyr-204) and anti-phos-
pho-STAT3 (Tyr705) were purchased from New Eng-
land Biolabs (Frankfurt am Main, Germany, f.c. 1:200
and 1:100 respectively). Anti-Erk1/2 (pan) was pur-
chased from BD Bioscience (Heidelberg, Germany, f.c.
1:500) Anti-EGFR and an EGFR-C-terminal peptide
were purchased from Abcam (Cambridge, UK, f.c. Anti-
EGFR 1:500).
Alexa-594-labeled chicken anti-rabbit IgG and Alexa-
633 goat anti-mouse IgG were purchased from Molecu-
lar Probes/Invitrogen (Karlsruhe, Germany; f.c. 1:1000
for immunocytochemistry, 1:500 for immunohistochem-
istry). FITC-coupled anti-mouse IgG was purchased
from Dianova (Hamburg, Germany; f.c. 1:500 for immu-
nocytochemistry and 1:200 for immunohistochemistry).
FITC labeled IB4 (f.c. = 1 ng/μl) and unlabeled IB4 (f.c.
100 ng/μl) were purchased from Sigma (Taufkirchen,
Germany).
DRG-cultures
Cultures of dissociated DRG were prepared from male
Sprague Dawley rats as described previously [31]. The
rats were killed by CO2 intoxication and L1-L6 DRGs
were removed, desheathed, pooled and incubated with
collagenase (final concentration (f.c.) 0.125%; 1 h, 37°C).
The neurons were dissociated by trypsin digestion (f.c.
0.25%, 1176 u, 8 min, 37°C) and triturated with a fire-
polished Pasteur pipette. Axon stumps and dead cells
were removed by centrifugation (5 min, 100 g). Viable
cells were resuspended in 12 ml of NeurobasalA/B27
medium, plated 0.5 ml/culture onto polyornithine/lami-
nin-precoated glass coverslips (12 mm diameter), and
incubated overnight in 24 well plates at 37°C in 5% CO2.
Cell stimulation
After incubation for 15-20 h cells were stimulated with
the growth factors NGF, GDNF, EGF or pharmacologi-
cally with Phorbol 12-myristate 13-acetate (PMA),
respectively. To ensure homogeneous mixture of the sti-
mulants, a volume of 250 μl out of the 500 μl culture
medium was removed from the culture well, mixed
thoroughly with the stimulant, and added back to the
same culture. Negative controls were treated alike but
without the addition of any reagent. To reduce mechan-
ical cell stress the stimulus was added very slowly (250
μl in 6 s) using an automatic pipette (Multipette® pro,
Eppendorf). After treatment, the cells were washed once
with phosphate-buffered saline (PBS) and fixed with par-
aformaldehyde (4%, 10 min) at room temperature (RT).
For staining with the pSTAT3 antibody cells were fixed
and permeabilized with methanol (10 min, -20°C)
Immunocytochemistry
Paraformaldehyde-fixed cells were permeabilized with
0.1% Triton X-100 (10 min, RT), followed by three
washes with PBS (5 min, RT). After blockage of nonspe-
cific binding sites (5% bovine serum albumin (BSA) and
10% normal donkey serum in PBS; 1 h, RT), the cultures
were probed with primary antibodies against target pro-
teins (antibody concentrations against target proteins as
indicated in Methods, Antibodies section) in 1% BSA in
PBS (1 h, RT), washed three times (1% BSA in PBS; 5
min, RT), and incubated with secondary antibodies (1 h,
RT). After three final washes (PBS; 5 min, RT), the cul-
tures were mounted with Fluoromount-G (Southern
Biotech⁄ Biozol) containing DAPI (0.5 μg/ml). Staining
with the isolectin IB4 from Bandeiraea simplicifolia was
performed in a solution containing 0.1 mM Ca2+, 0.1
mM; Mg2+ and 0.1 mM Mn2+ in PBS (1 h, RT), fol-
lowed by three washes (0.1 mM Ca2+, 0.1 mM; Mg2+
and 0.1 mM Mn2+ in PBS, 5 min, RT) before mounted
onto microscopy slides with DAPI/Fluoromount-G.
Immunohistochemistry
Prepared L3-L6 DRGs were fixed with paraformaldehyde
(4%, 1 h, RT), washed 3× with PBS (20 min, RT), satu-
rated by increasing concentrations of sucrose (10, 20
and 30%, each saturation step overnight, 4°C) and
embedded in Tissue tek® (EMS Science Services). Mate-
rial was frozen on dry ice and stored at -80°C. DRGs
were cut in 20 μm sections using a Cryostar Cryostat
Andres et al. Molecular Pain 2010, 6:98
http://www.molecularpain.com/content/6/1/98
Page 3 of 16
Page 4
HM560 and dried on microscope slides (30 min, 37°C).
The sections were fixed by paraformaldehyde (4%, over-
night, 4°C). Immunostaining was performed as indicated
in the “Immunocytochemistry” section.
QuAM
Cells were evaluated with a Zeiss Axioplan 2 microscope
controlled by the software Metacyte (Metasystems).
Images of 1280 × 1024 pixels were taken using a 10×
objective. The exposure time was defined automatically
so that maximal 1000 pixel/100 μm2 reached saturation
(i.e. maximal intensity values), but was maximal 0.96s.
For automatic neuron recognition the following para-
meters were defined: size (150 -1500 μm2), form (aspect
ratio = 2; concavity depth = 0.25), contrast (object
threshold 30%). The integrative pixel intensity of each
selected neuron was normalised against the respective
neuron area and exposure time. The influence of varying
exposure times on the resulting intensity value is insig-
nificant (Additional File 1: Figure S1). For cell identifica-
tion the neuron specific PGP 9.5 immunostaining was
used as independent selection marker. Fluorescence
intensities derived from phospho-Erk1/2 antibody and/
or IB4-signals were quantified on independent color
channels.
Random sampling
Our single cell measurement data are not distributed
normally as they neither pass a normal distribution test
nor a chi-square test. Thus, statistical comparisons of
untreated and treated cultures were performed using a
random sampling approach. “Virtual culture wells” were
created by randomly picking 250, 1000, 5000, 10000 or
20000 cells out of all measured cells and the average
intensity of these virtual wells was computed. Sampling
and computing the average was repeated 10,000 times.
One and the same cell-value could be picked more than
once. The sample mean is, by the Central Limit Theo-
rem, approximately normally distributed, with mean
equal to the population mean.
Confocal microscopy
For evaluation of protein expression confocal images of
cells and sections derived from DRG were taken on an
inverted Zeiss LSM 510 Meta or a Zeiss LSM 700 with
63× or 40× objectives. Plasma membrane staining was
analysed via intensity-histograms with the software Ima-
geJ by measuring the fluorescence intensity profile along
a cell crossing line. The intensity of nucleus localized
pSTAT3 signal was quantified also with ImageJ.
Evaluation of pSTAT3 translocation
Cells were evaluated with a Zeiss Axioplan 2 Imaging
using a 63× oil-immersion objective. Fifty randomly
selected cells per culture were evaluated. Data are
plotted as mean percentage of translocating cells per
evaluated cultures + SEM based on the number of eval-
uated cultures. All counting was done by the same
observer. All treatments have been repeated three times
with DRG-neurons from different rats.
RT-PCR
For RNA extraction DRGs or brain from male Sprague-
Dawley rats were isolated. Total RNA was extracted
using the Nucleospin RNA/Protein Kit (Macherey-
Nagel, Düren, Germany). DNA was digested for 15 min
at 37°C with RQ1 DNAse (Promega, Mannheim,
Germany). RNA was precipitated (0.1 volume 3 M sodiu-
macetate, 2.5 volumes ethanol; 20 min, 20°C and 10 min,
20800 g), washed with 70% ethanol (5 min, 20800 g), dried
and resuspended in RNAse-free water. cDNA was created
using SuperScript III First Strand Synthesis SuperMix for
qRT PCR (Invitrogen, Karlsruhe, Germany). An EGFR-
specific fragment was amplified with specific primers
(5’- CATCCAGTGCCATCCAGAAT - 3’ (forward) and
5’- CTTCCAGACCAGGGTGTTGT - 3’ (reverse)) via
PCR (2 min at 94°C, 30, 35 or 40 cycles of 30 s at 94°C, 30
s at 53°C, 30 s at 72°C and a final step of 10 min at 72°C).
The PCR product was analysed on a 2% agarose gel and
imaged with a Herolab EASY 440K gel documentation
system.
Testing of mechanical nociceptive threshold
The nociceptive flexion reflex was quantified using a
Randall Selitto paw pressure device (Analgesymeter;
Stoelting, Wood Dale, IL), which applies a linearly
increasing mechanical force to the dorsum of the rat’s
hindpaw. The nociceptive mechanical threshold was
defined as the force in grams at which the rat withdrew
its paw. The protocols for this procedure have been
described previously [32,33]. All experiments were per-
formed at the same time of day. In the week preceding
the experiments, rats were familiarized with the testing
procedure at 5 min intervals for a period of 1 h per day
for 3 days. Baseline paw withdrawal threshold was
defined as the mean of six readings before test agents
were injected. Each paw was treated as an independent
measure, and each experiment was performed on a
separate group of rats. Each group of rats was treated
with only one agonist injected intradermally. The noci-
ceptive threshold was measured 30 min after the admin-
istration of the respective reagent. The reagents (see
description in the Results section) were injected as
described previously [34,35].
Statistical Analysis
For statistical comparisons of EGFR fluorescence inten-
sities and pSTAT3 nucleus localized fluorescence
Andres et al. Molecular Pain 2010, 6:98
http://www.molecularpain.com/content/6/1/98
Page 4 of 16
Page 5
intensities the Students t-test with Welsh-correction was
applied. Statistic analysis of pSTAT3 nucleus transloca-
tion as well as comparison of pErk1/2 and Erk1/2 fluor-
escence intensities between IB4(+)- and IB4(-)-neurons
was done with an one-tailed unpaired t-test. For statisti-
cal comparison of NGF and GDNF treated vs. untreated
cultures the sample mean resulting from the random
sampling approach were tested by an one-tailed
unpaired t-test. Statistic analysis of behavioral data was
done with one-way ANOVAs followed by Dunnett’s test
for comparisons with one control value. P < 0.05 was
considered as statistically significant.
Results
Software-based identification of dissociated DRG-neurons
The heterogeneity of primary DRG-neuron cultures
challenges the first step of a quantitative image analysis,
the object identification. The DRG-neurons, but not
axon stumps or glia cells have to be identified by setting
suitable “object identifier” parameters in the image-
analysis software. Images of immunofluorescently
labeled dissociated DRG-neurons culture were taken by
an automated microscope (Figure 1A) and neurons
identified by the image analysis software “Metacyte”
(Metasystems). Neurons were visualized with an anti-
body against the neuronal marker PGP 9.5. Object iden-
tification parameter optimization was performed and
allowed the identification of neurons based on object
size (150-1500 μm2), roundness (ratio of longest to
shortest diameter < 2), smoothness of cell perimeter
(concavity depth < 0.25), and relative PGP 9.5 immuno-
fluorescence intensity (> 30% difference to the sur-
rounding) (Figure 1B).
In the culture 16 ± 1% of all cells identified visually as
neurons based on their shape and neuronal marker PGP
9.5 expression were in cell clusters. These clustered cells
were rejected as cells function very differently if cultured
as separate cells or connected to others. Further 3 ± 2%
of all neurons were cut by the viewfield borders of the
images and were therefore also rejected. Of the remain-
ing 81% of all neurons in culture, 93 ± 2% were detected
by the software indicating a very robust and well opti-
mized detection algorithm.
Single DRG-neuron based studies require large number of
analyzed cells per treatment
One signalling component important in nociceptor sen-
sitization is the MAP-kinase Erk1/2 [36]. Activation of
Erk1/2 is commonly detected by use of antibodies speci-
fic for the Thr-202/Tyr-204 phosphorylation site of
Erk1/2. We tracked the basal phosphorylation Erk1/2
signals in untreated cultures with basal levels of activa-
tion and investigated 49,502 single cells. The degree of
immunofluorescently detected phosphorylation shows
an intensity distribution with a standard deviation of ±
180%.
Considering the high standard deviation of intensities
we next determined, how many cells have to be mea-
sured to describe the distribution sufficiently and what
level of sensitivity is reached thereby.
One can estimate, how good the “real” distribution is
detected by comparing the distribution of culture wells
with small numbers of cells with the distribution of a
very large number of measured cells. To do so, out of
49,502 measured unstimulated cells we randomly
sampled 10,000 artificial cultures consisting each of 250,
1000, 5000, 10,000 and 20,000 cells per virtual well,
respectively (Figure 1C). As expected, sampling more
cells per culture resulted in an ever-narrowing degree of
deviation of mean-culture-intensities. For 250 cells the
standard deviation of the mean is +/- 11%, for 1000 +/-
6%, for 5000 and 10,000 cells +/- 2%, and for 20,000
cells +/- 1%. This indicates, that mean culture intensity
changes below 22% (two times standard deviation)
intensity change cannot be assumed statistically sound if
only 250 cells are investigated. Of note, an intensity
increase in only a subgroup of e.g. 1/5 of the cells,
requires with 250 measured cells per culture an increase
of at least 5 × 22% = 110%, i.e. roughly a doubling of
the intensity, within the subgroup to result in an overall
mean change of 22%. But investigation of more cells
allows the detection of smaller changes even down to
some few percent change (standard deviation for 10,000
measured neurons = +/- 2%) indicating high sensitivity
of our relative immunofluorescence approach despite
the apparent huge heterogeneity of DRG-cultures.
Subgroups can be defined by classical expression
markers
In most published reports of immunofluorescently-labeled
marker-protein-based subgroups, the categorization into
“marker-positive” and “marker-negativ” is performed by a
trained experimenter judging by eye [2,3,37]. This assumes
a clear distinction between the positive and the negative
cells. This assumption nevertheless has never been tested
thoroughly.
One well established marker for such a DRG-neuron
subgroup is the isolectin B4 (IB4) identifying preferably
non-peptidergic GDNF-dependent nociceptive C fibers.
Thus we measured the intensities of single cells labeled
with IB4. The intensities showed a continuum of intensi-
ties (Figure 1D). Thus the assumption of clearly separated
intensity ranges for marker-positive and marker-negative
cells is not valid.
Nevertheless, representing the data by an intensity his-
togram one is tempted to differentiate one population in
the lower intensity range (large peak on the left) and a
shallow but widespread distribution with much higher
Andres et al. Molecular Pain 2010, 6:98
http://www.molecularpain.com/content/6/1/98
Page 5 of 16
End of preview.
