Gene expression in the inner ear - IGBMC

shrubberystatuesqueData Management

Dec 1, 2012 (4 years and 8 months ago)

300 views

1



Sommaire

Introduction

................................
................................
................................
................................
.............

2

Material and Methods

................................
................................
................................
.............................

4

Annotati
on

................................
................................
................................
................................
.......

4

Data management and web site

................................
................................
................................
.....

4

Visualization, querying and analysis tools.

................................
................................
......................

4

Dendogram of tissue correlation
................................
................................
................................
.....

5

Interactomics

................................
................................
................................
................................
...

5

Gene Ontology
................................
................................
................................
................................
.

5

Overview of the expression patterns

................................
................

Error! Bookmark not defined.

RESULTS AND DISCUSSION

................................
................................
................................
......................

6

Hierarchical tissue clustering

................................
................................
................................
..............

6

Overview of gene expressions in the sensory organs

................................
................................
.........

7

a
-

Gene expression in the inner ear

................................
................................
...............................

8

b
-

Gene expression in the sensory retina (R)

................................
................................
.................

9

c
-

Gene expression in the olfactory organ (O)

................................
................................
...............

9

d
-

Gene expression in the vibrissae follicles (V)

................................
................................
.............

9

Network analysis of gene expressed in the five sensory epi
thelia (KUROV)

................................
......

9

Gene functions in relation to Gene Ontology (GO) annotation

................................
........................

11

Network analysis of genes expressed in the olfactory organ

................................
...........................

11

Gene expression at the onset of olfactory function

................................
................................
.........

12

Human known diseases related to genes found in KUROV

................................
..............................

13

CONCLUSION

................................
................................
................................
................................
.........

14

ACKNOWLEDGMENTS

................................
................................
................................
...........................

15

REFERENCES

................................
................................
................................
................................
..........

15

FIGURE LEGENDS

................................
................................
................................
................................
...

20




2



I
ntroduction

Stages of development are governed by several cell processes that are mainly under the control of gene
networks. One of the present challenge
s

of developmental biology is to decipher specific pathways involved
during different phases of an organism ontogenesis. Generally each stage of development is controlled in part by
a network of genes that are often specific to an organ or even a tissue and

since the genetics pool of most
organisms reach several thousands of genes, it was
not easy

until the recent past to decipher the regulatory
pathways

and interplay
. Thanks to the improvement of genetics technologies that provide high
-
throughput data
and l
arge
-
scale analysis of gene expression along with tissue
-
specific and whole organ cDNA libraries with data
mining techniques aimed to extract useful information (). For example m
icroarray technology has been
successfully applied to profile the expression l
andscape of the entire transcriptome and to get a global view of
genes function in different organs
(
Stansberg et al., 2007
).

However this powerful technique very often lacks the
desirable fine resolution necessary for understanding the molecular mechanisms operating

at the tissue or/and
cellular levels. In order to get a better relationship between genes and anatomical structur
es with high
-
throughput
yield, a semi
-
automated

in situ

hybridization (ISH) technology associated with high resolution digital imaging
has been developed
(Alvarez
-
Bolado and Eichele, 2006; Diez
-
Roux et al., 2011; Lau et al., 2008; Magdaleno et
al., 2006; Neidhardt et al., 2000; Ng et al., 2009; Visel et al., 2007)
. ISH provides
in addition

a semi
-
quantitative
estimate of mRNA abundance; however, it allows to pin
-
point transcript distributions at the cellular level within
anatomical substructures

(Gofflot et al., 2007)
. Several species are used as models for developmenta
l studies.
Among mammals,

genetic analyses have shown the high levels of similarity between the human and mouse
genomes where approximately 99% of mouse genes have homologues in the human genome
(
Landers et al.,
2001; Waterston et al., 2002
; Carninci et
al;, 2007
).

The well described relationships between mouse and human

genomes imply that whenever genes are involved in development or pathology, their homologues usually exist
and
can be

identified in the mouse genome (
Nguyen & Xu, 2008
; Kitsios et al., 20
10
).

The present study aimed at characterizing gene expression patterns in five

developing sensory organs, at a
specific stage of mouse development
-
embryonic day E14.5
-

using the ISH data generated by the GenePaint and
EURExpress projects. This stage ha
s been chosen (see
Diez
-
Roux et al., 2011)

because most organs are already
morpho
lo
gicall
y differentiated, although specific cell
-
types

may not beat the same stage of differentiation in
various organs.

W
e
have
analyzed 2000 transcripts chosen at random

from the GenePaint

(
http://www.genepaint.org
)

/
EURExpress (
http://www.eurexpress.org
)
ISH
collection
.
Twenty
-
five
tissues

from
the middle ear, inner ear, retina, brain, olfactory organ,
bones
and
vibrissae

follicles
were analyzed in detail, and
the
expression pattern
s were annotated

and illustrated

including more than 11

000 pictures of the
2000 transcript
expressions
.

T
o
give
access
to the biologists
to

an integrated system

allowing the analysis of this information
we developped

ImAnno, as a new, web
-
accessible and searchable database. ImAnno (
http://lbgi.igbmc.fr/imanno
) offers
the
possibility to the biologists to upload new images as well as links to images hosted by other web sites (as
GenePaint or Eurexpress), to annotate them using user defined formulars and stores this information in a
dat
abase allowing

visualisation and querying
. It provides a powerfull search tool which permits the user to
create and save complex queries, to obtain lists of genes as for example ‘expressed in a set of tissues and strictly
3



absent in other’ and furthermore

to combine them with other sources. ImAnno provides also various integrated
analysis tools such as interactomics, gene ontology, transcriptomic expression, mutation, pathology, etc.

These bioinformatic tools

allowed us to
stud
y

the possible genetic relat
ionships of the different tissues analyzed,

as well as to find out some interesting xxxxx
Among the five sensory organs, two
-

the inner ear and the
olfactory organ
-

derive from sensory placodes, the retina is formed from the anterior region of the brain,

and the
vibrissae

follicles from placodes of the muzzle skin.

Our analysis was focused

on the specific sensory region of
five
presumptive sensory organs, i.e. Kölliker’s organ (K) for the cochlear canal, the utriculus sensory region (U)
for the vestibular
organs, the sensory retina (R) for the eye, the sensory region of the olfactory organ (O) and the
vibriss
a
e follicles

(V) for skin mechano
-
receptors.
Hereafter w
e use

the following acronym for these 5 sensory
organs
:

KUROV
.

Comparison between gene expression patterns has been undertaken with
the
bioinformatics
tools
provided by ImAnno
in order to find any similarities between patterns of expression, and highlight
pathways operating during development of these sensory organs.
After analyzing the set of genes expressed in
the 5 developing sensory organs (KUROV), we concentrate on one of these organs
-

the olfactory organ
-

as an
example for further genetic interaction analysis. Finally we discuss the relationship between gene ex
pression
patterns and known pathologies in human.



4



Material and Methods

ImAnno is a web based integrated system allowing the annotation, management, visualisation and querying of
the gene expression information of the
in situ

hybridisation images.

Annota
tion

A first interactive tool provided by the web site allows the user to proceed to the annotation of the genes. For that
the user is invited to define a list of tissues he is interested in. This tissue list is specific to each project : 20
tissues for th
e Eye annotation (ie retina outer layer, lens anterior epithelium, mesenchim eyelid, …), 36 for the
Teeth

(ie oral epithelium, molar mesenchymal compartment dental sac, incisor gubernaculum, …), 25 for the Ear

a
nd

SensorySytem (see tissue list table TisLi
s). The annotation consists of checking each tissue for an
expression value within Negative, Weak, Medium, Strong or NotAvailable

(
according

to
http://www.genepaint.org
;
Visel et al., 2004)
. The annotation form

(see

Fig
-
AnnotForm)

accepts also a free text
input for each tissue as well as a general free text for the whole annotation. An additional «

tissue

» 'General
expression' is also proposed.

Notate that o
ne annotation form is associated to one gene. The aim of a gene annotation is firstly to define the
gene, then to choose a set of
in situ

hybridization images which show the expression of this gene within the
tissues and final
l
y to estimate and check the e
xpression level for each tissue. A gene can be defined by its gene
name but a better and more powerful method is to

copy
-
paste the gene metadata
provided by
the
GenePaint web
site

(http://www.genepaint.org
), facilitating therefore the detection and lin
k to

the corresponding images. Several
images
from different sections with different magnifications are often necessary to clearly show the expression
level. The images are either simple http links to the visualization facility provided by GenePaint, Eurexpres
s or
any other web site, or locally hosted images uploaded by the user himself.

For our analysis we used following protocole :

T
he probes were obtained from

GenePaint w
h
ere sequence of the
template used for in vitro transcription o
f the RNA probe can be o
btained
. The automated obtention of non
-
radioactive ISH from E14.5 mouse cryosections has already been described in detail, (Carson et al., 2002; Visel
et al., 2004, 2007). Gene expression patterns were digitally photographed

at IGBMC

by a DMBL Leica
micro
scope equiped with a Photometrics camera with the CoolSNAP software

(v. 1.2)
. The
images were
deposited on the
ImAnno database and are publicly available. Expression patterns for all

tissues were manually
annotated by the first author and checked for valid
ation one year after the initial observation.


For each form submission the new annotation is integrated in a relati
onal database, storing values,
ownership and
history of the annotation act. The same user or any other authorized annotator can re
-
annotate
any tissue as long
as the annotation of the gene is not marked as 'Approved'.

Data management and web site

Information about the genes, there corresponding images, the
per

tissue expression level annotation as well as
the free text comments are stored in a

Po
stgresql relational database and can be queried through the ImAnno
website. The web site (PHP, HTML, Javascript) provides user authentificat
ion with adapted read and write
access rights.

Visualization, querying and analysis tools.

Any authorized user,
according to his access rights, can query and visualize all or only some parts of the
information from the database. Looking for a gene, he gets access to the images and to the expression
annotations. ImAnno offers an explicit visual display highlighting t
he expression on a synthetic picture showing
the tissues colored according to their level (grey: NotAvailable, blue:Negative, yellow:Weak, orange:Medium,
red:Strong).

ImAnno provides also a powerful querying system. Searches can be done by

genename

and more
interestingly
by the
per

tissue expression level. For example it is possible to search all genes for which there is a strong
expression in tissue T1 and medium or weak in tissue T2. These kind of select patterns can be easily combined in
what we c
alled “sieves”

(Fig
-
Sieve)
. A sieve is a boolean
combination (with and & or) of
such atomic patterns.
ImAnno provides an easy tool to create and save these sieves allowing therefore any user to construct powerful
pattern searches, from scratch or modifying

existing one. Furthermore the lists of genes obtained by several
5



sieves can be combined as logical union, intersection and/or complement operations using a dynamic html
formular which can be modified and extended by the user.
This possibility
was widely u
sed

by the biologists t
o
define more and more complex patterns as
for example
KUROV (present in the five tissues),

or, using lowercase
letters for absence,

kuRO (absent in K and U but present in R

and

O)
, and combining the lists,

O
_notKUROV
(
present in

O

but not present

simultaneously in the five other tissues
,
in other words present in
O

but absent in at
least one other organ

(same as
O
_not
K
URV)
).
Table Tab
-
Repartition2000
shows the repartition of the 2000
genes within the 32 (=2^5) po
ssibilities

involvi
ng the five tissues (it confirms the proximity of ROV, ).
Finally
the web site offers many integrated tools which can be applied to these lists of genes such as Gene Ontology,
Interactomic (using the String database and Cytoscape), known phenotypes, relati
on to diseases, comparison
with externa
l lists, etc. It allows also calculating and displaying
correlations between tissues, sieves and any
combination of them.

Dendogram of tissue correlation

To obtain the dendogram we first compute a distance matrix be
tween all tissues taking into account the
expression values of the 2000 annotated genes, using a Spearman’s rank correlation coefficient with following
numeric values 0:negative, 2:weak, 3:medium and 4:strong (R function
cor(.., method=”spearman”)
).
This d
istance matrix is used by the program FastME (ref Gascuel O.) to construct the phylogenic tree.


Interactomics

Protein
-
protein interactions were obtained from the STRING database containing known and predicted physical
and functional protein
-
protein intera
ctions. STRING was used in protein mode, and only interactions with high
confidence levels (>0.7) were kept.
Interactomic networks were visualized
with the Cytoscape software to
facilitate the analysis of the interactomic networks, which may be very huge. Three different
“zooming in levels”
were created
:
the
first level

deals only with initial genes with at least one interaction (primary network), th
e
second level

describes the initial genes and only the genes from the STRING database exhibiting interactions
with two initial genes. The
third level
display the genes from the initial list (initial genes) and all the genes
exhibiting at least one interac
tion with any initial gene found in the STRING database.

Gene Ontology

The gene ontology database from
http://www.geneontology.org


David

Copy paste
gene list in the David web site

Morbide Map

ImAnno allows to check the presence in
distant databases
as MorbideMa
gscope





6




RESULTS AND DISCUSSION

Overview of ISH expression


Concerning the different level of ISF expressiona

significant percentage (16
.
5%, i.e. 331) of the genes
analyzed did not present any ISH signal in the 25 tissues investigated. In some specific tissues the absence of
signal was from, for example,
73.9% for the stria vascularis from the inner ear to only 41.9% for the retina.

Several

raisons may be put forward to explain the absence of ISH signal: firstly, the gene was not expressed
at all then, obviously did not generate a signal. This is confirmed by the observation of ISH data from the
whole embryo from over 18000 transcripts inclu
ding our 2000 probes where 18% were not detected
(Diez
-
Roux et al., 2011)
. Our 16.5% of negative results is close to the 18% obtained for the whole embryo
confirming that our sampling of 2000 probes
.


The majority of transcripts tested presenting an ISH signal showed a weak expression as summarized in
Table 2. Although, this results is not surprising since a serial analysis of gene expression (SAGE) showed
that most transcripts (86%) were expressed by l
ess than five copies per cell
(Zhang
et al., 1997
)
. The main
problem was to decipher the weak ISH signal from background. The level o
f gene expression was controlled
for every batch of genes (15 genes per batch). The weak expression was controlled by using the genes
Rar

2

that presents a good example of low transcript expressions in several organs (Mollard et al., 2000). The level
of ba
ckground was checked by omitting the probe. Sections were double checked at least one year apart
without much difference for the annotation.


Hierarchical tissue clustering

We established a tissue dendrogram, based upon a hierarchical clustering
of 25 dif
ferent tissues
using
average linkage of the differentially expressed genes according to our analysis (Fig. 1). This clustering

revealed interesting relationships between the different tissues analyzed
. Three main nodes can be
distinguished: node A is related to genes expressed in tissues mainly derived from the inner ear. Node B is
more heterogenous, although the expression patterns cluster for a large part in structures derived from
mesenchymal tissue
s, such as different types of cartilages or bony structures (middle ear ossicles), secretory
organs (the stria vascularis), the choroid plexus, and other mesenchymal tissues from the inner ear and
middle ear. Node C is mainly composed of genes expressed in

nervous tissues and ectodermal/mesenchymal
tissues.

Genes from node A1 are expressed in the sensory region of the utricule/saccule, the non
-
sensory region of the
utricule/saccule, and the semi
-
circular canals
, (Fig. )
. All these structures are derived fr
om the otic placode
epithelium of ectodermal origin, and the co
-
expressed genes may present close functional relationships, e.g.
genes encoding balance receptors (
Cristobal et al., 2005; Sajan et al., 2007
). Genes expressed in tissues
7



derived from sub
-
node

A2
are very close and
belongs to the cochlea such as the inner spiral sulcus and
Kölliker’s organ that are devoted to hearing function.
Two
other tissues form the cochlear canal, the stria
vascularis and the outer spiral sulcus, are present in another nod
e (node B1) that also comprises the
endolymphatic organ.
The origin
of these tissues f
rom the same node (B1)
may be explained by the fact that
most cells of these three organs are of mesenchymal origin and later will be involved in
different
fluid
transpor
t and ion secretion
(Steel and Barkway, 1989
; Minowa et al., 1999; Couloigner et al., 2004
)
.
It is
interesting to note that the choroid plexus of the 4th ventricle, whose function is mainly related to the transfer
of molec
ules important to the brain during development
(Emerich et al., 2005)
, belongs to another sub
-
node
(B2). This organ is made of a single epithelial cell layer originating from the neural tube, whereas its
mesenchymal core originates from paraxial mesoderm
(S
wetloff and Ferretti, 2005)
. The tissues from sub
-
node B2b, despite their various embryological origins, may have some common mesenchymal tissue
characteristics. The otic capsule is part of the chondrocranium, while the middle ear ossicles are from the
splanchnocranium deriving from the branchial arches
(Depew et al., 2002)
. These two structures cluster with
the cartilage primordia of rib reflecting similarities in gene expression relating to chondrogenic
differentiation. Node C is made of two sub
-
nodes, where sub
-
node C1 is
composed of nervous organs such as
the olfactory organ, the hindbrain, the sensory retina and the stato
-
acoustic ganglion that are genetically very
close despite their different embryological origin
(Rossant and Tam, 2002)
. This proximity may be explained
by their similar genetics pathways acting dur
ing development, especially for genes controlling neural
differentiation.
However, the olfactory organ is well differentiated from the three other

organs,

this can be
explained by the fact that we included in this organ, expressions from the lamina propria

that includes several
different cell types
(Kessel and Kardon, 1979)
, increasing the genetics heterogeneity

of this sub
-
node
.
The
second sub
-
node (C2) comprises the vibrissae follicles of ectodermal origin, the external acoustic meatus
which is directly formed from the

ectoderm of the first branchial cleft, and the tympanic membrane made of a
fibrous layer of mesodermal origin surrounded by two epithelia layers
(Mallo, 2003)
.

This analysis by hierarchical clustering shows that grouping of different tissues corresponds either to a
common embryological origin like for the node A with the tissues from

the inner ear, or their prospective
function like for nervous tissues (node C1). Other tissues may be more related to their tissues of origin such
as the mesenchyme for node B1. Another interesting observation is that our sampling of 2000 genes mimics a
f
unctional aspect or an embryological origin that reinforces our random sampling of our 2000 genes.



Overview of gene expressions in the sensory organs

In the following section,
we

focused on genes expressed in
each of
the five sensory epithelia
(designated by
the

acronym KUROV, where K stands for Kölliker’s organ, U for the sensory region of the utricle, R for the
retina, O for the olfactory organ, and V for the vibrissae follicles.
Out of 2000 transcripts

analyzed, 1424

presented an ISH signal
i
n at least one of the tissues of these

sensory organs (
Supplementary
Table

S
1) and
233 showed a ubiquitous distribution in all tissues.


8



a

-

Gene expression

in the inner ear

The inner ear is made of two sensory organs
: the cochlea and the vestib
u
le
. The cochlea
consists of a
complex bony canal that makes two and half

spiral turns around a

central

axis. This organ is
devoted

to the transduction of acoustic signals
. The
vestib
u
le is made of 5
vestibular
organs
: 3 cristae from the
semi
-
circular canals
and

2

maculae respectively fr
om the saccule and the utricule. These vestibular
receptors are involved in the balance.

The basal cochlear canal (K)

In the cochlea at E14.5, the ventral region of the basal cochlear canal mainly includes
the
Kölliker’s
organ
,

wich

includes
the prospective sensory region

(Figs. 2A & B)
. Out of 2000 genes tested, 784
(39.2%)
presented an ISH signal
in this structure
(Tables 2 & 3
, Fig. 3A
) where some could
show

a
large expression like
Ctgf
, (Fig.
2
A), or other a very restricted pattern like
Shc3
, (Fig.
2
B).
Thirty five
genes from the Kölliker’s organ have been found to be up regulated in a
high density expression
microarrays study of the cochlea at E14.5 (Sajan et al., 2007), such as
Gsta4, Rcn1 and

Wnt7a

presenting a strong or medium expression in the cochlea from ISH.
Gsta4

has been found in the stria
vascularis in adult mice in relation with melanocytes (Uehara et al., 2009), while our data present a
strong expression in the sensory region of the
cochlea suggesting maybe another function for this gene
during development.

Only 9 genes were specific to
the
Kölliker’s organ compared to the four other
sensory tissues (Table 3)
. For instance one of them,
T
LR2

(Toll like receptor2) is involved in the inner
ear inflammation secondary to chronic middle ear otitis media
where
a
monocyte
-
attracting chemokine
is up
-
regulated in the spiral ligament fibrocytes
of the cochlea
(Moon et al., 2007)
.

The utricular sensory

region (U)

The vestibular

organs
were

not studied in the same detail as the cochlear canal, because very often not
all 5 vestibular organs were present on the histological sections hybridized for a given probe. In our
annotation system, when an ISH signal

was observed in one vestibular organ, it was scored positively
for the vestibular receptors. This annotation will not reflect the exact differential distribution of some
genes in the five vestibular organs, as it has been shown by an extensive microarray
study of the

developing sacculus and utriculus
(Sajan et al., 2007)
.

From our observations, 876 (43.8%) of the 2000 genes analyzed presented an ISH signal in the
prospective vestibular
receptors, the supporting cells, and maybe some neighboring cells such as
transitional cells

(Tables 2 & 3
, Fig. 3A
). Several genes showed a particularly strong

expression in the
utricular

epithelium
, (Figs.
2
C & D)
:

Forty four genes from the sensory regio
n of the utricule have
been found to be up regulated in a high density expression microarrays study from the utricule at
E14.5 (Sajan et al., 2007), such as
Cldn6, Eps8l2, Grp and Rcn1

all
presenting a strong
ISH
expression in the utricule.
Cldn6
has been

found to be strongly expreessed

in the otocyst in the E9
,
5
-
E10,5 mouse embryo
(Anderson et al., 2008
),
and in the endolymphatic sac epithelia of adult rat
suggesting a
role in the
ions regulation
of

the inner ear, (
Matsubara et al., 2011).
Only 11 genes were
specifically present in this sensory region
-

i.e. were not expressed in the other four sensory epithelia
(Table3).
One of them
Oc 90

is present very early in the murine inner ear during ontogenesis and is
involved in the formation of ot
onia that are involved in the balance function, (Thalmann et al., 2001).

Focusing on genes expressed in the sensory epithelia (i.e, Kölliker’s organ in the cochlear canal, and
vestibular sensory epithelium from the utriculus), out of 784 genes found in th
e Kölliker’s organ, and
876 in the utriculus sensory region, 703 genes were common for both sensory regions
, (Fig. 3
B; Table
3).


9



b
-

Gene expression

in the sensory retina

(R)

In the retina,

out of 2000 genes analyzed, 1162 transcripts (58%) were
expressed in the sensory region (the
epithelium of the neural retina
;
see
Table
S
1 for a complete list). From these genes, the large majority exhibited a
weak expression (62.3%) and only 5.2% of the expressed transcripts showed a strong signal (Table 2
, Fi
g. 3A
).
Several genes were strongly expressed in the sensory retina and very often in nervous tissues like the hindbrain,
the stato
-
acoustic ganglion and the olfactory organ
like
Mid1

and
Fubp1

(Figs2 E & F).
Several of these ISH
gene expressions have not
been described in the retina

during development, (Blackshaw et al., 2004)
. Other have
been found in the retina like

like
Crabp2

expressed at E13.5 (Ruberte et al., 1992), however its deletion did not
produce a specific phenotype in the retina, (Gorry et al., 1994).
From the pool of probes tested in the retina and
those observed on the four other sensory regions, 71 genes were spec
ifically expressed in the sensory retina
(Tables 3
)
.
730
genes
were common between
the
Kölliker’s orga
n and the sensory retina (Fig. 3
B; Table 3).


c

-

Gene expression

in the olfactory organ
(O)

Rodents possess multiple organs for detecting odors: the olfactory epithelium, the vomeronasal organ, the
Grueneberg ganglion and the septal organ
(Breer et al., 2006)
. Here we report ISH expression restricted to the
olfactory organ with its sensory epithelium, supporting cells and below the lamina
propria including blood
vessels, connective tissues, nervous fibers and Bowman glands
(Kessel and Kardon, 1979)
. ISH signals were
observed for 1175 probes in the olfactory organ, representing 59% of the tested transcripts
, (Fig. 3A)
. Some
genes were only expressed in the

prospective olfactory epithelium

like Gp38 (Fig. 2G), or in both the
olfactory epithelium and the lamina propria (Fig. 2H)
.
Seventy four

genes whatever the level of expression,
showed an expression presents only in the olfactory organ compared with the fo
ur other sensory regions
(Tables 3). These included
only
one

olfactory receptor
gene
, the

Olfr568
.
A large number of genes from the
olfactory organ were common to the retina
like the
Olfr 65

and in a decreasing number with the vibrissae
follicles and the
Kölliker’s organ, (Fig. 3B).

d

-

Gene expression

in the
vibrissae

follicles

(V)

From all transcripts
analyzed
, 1091 (54.5%) presented an ISH signal

(Tables 2 & 3
, Fig.
3A
).
Some genes
had
a strong expression
to the whisker follicles

such as
Mif

(Fig.
2I)
.

Another gene like

Tomm20

has been
found by a microarray study in the human head hair follicles, (Kim et al., 2006).

Fifty six genes were specific
to the vibrissae follicles (Table 3). Very often expression in the whisker follicles was related to an ISH sig
nal
in the skin. One of them
Hoxc13

was very restricted and only observed in the follicle.
It has been shown to
be essential for proper hair shaft differentiation in human (Jave
-
Suarez et al., 2002), as
Hoxc13

gene
-
targeted mice completely lack external hair and nails (Potter et al., 2011)
.



Network analysis of
g
ene

expressed in

the five
sensory
epithelia

(
KUROV
)

Further analyses of relevant genetic pathways and gene functions were carried out using bioinforma
tic
approaches
.
We first focused on genes expressed
all together
in

the five sensory epithelia

as designed as
KUROV.
There were

623 genes expressed in all five sensory tissues belonging to KUROV
(for a complete
10



list, see Supplementary Table 4).

We used the STRING database in order to find possible interactions
between these genes. Out of 623 genes, only 163

showed a direct interaction
s

with genes found in the
STRING database where they can be grouped into
one

major network
and two smaller ones
(
Fig.
4
A
). The
first network is composed of
5 sub
-
ne
t
works connected by specifi
c genes we call “gene
-
bridges”.

For
example the first sub
-
network 1a mainly
composed of
ribosomal proteins and mitochondrial ribosomal
proteins
that
is linked t
h
rough the
second
-
sub
-
network 1b by
Eif2s1

gene (Translation initiation factor 2,
subunit 1 alpha)
.
This gene is
essential
for

preserv
ing

the integrity of the endoplasmic reticulum and to
increase insulin production to meet the demand imposed by a high
-
fat die
t.
(Bac
k et al., 2009).

This sub
-
network

contai
ns

genes involved in protein processing in the endoplasmic reticulum
,
endocytosis
,
a
poptosis
,
and a growth factor receptor
.
The

first
sub
-
network is also connected to
a third one through the
Rps27l

gene
(Ribosomal protein S27 like) where genes are linked to several cellular processes such as
the
focal adhesion
,
cytoskeleton regulation and neural development
.
This third

sub
-
network of ge
nes is related to a fo
u
rth

one
(1d)
through the
Aurora kinase A (
Aurka
)
gene.
Aurora

A

is an integral mitotic kinase essential for early
embryonic development and for microtubule dynamics of post
-
mitotic neurons that may be involved in
establishing cell polarity and axon/dendrite elaboration in young neurons

(Lef
kowitz and Gleeson, 2009)
.

In
this fo
u
rth

sub
-
netw
ork several prote
a
some genes were

present
, like

other genes involved in important
cellular pathways such as Wnt

and Hedgehog signaling pathways, cell cycle, meiosis, and Ubiquitin
-
mediated proteolysis
.

Csnk1a1
, encoding a casein kinase

involved in Wnt

and
Hedgehog signaling pathways

provides

the connection with the fifth

sub
-
network
(1e),
mainly composed

of genes
encoding tubulin
,
actin
subunits
,
centrosomal and/or cilia components
.


A
second

network independent from the previous one is made of genes mainly involved in transcription with
several spliceosome genes
,

polymerases and transcription factors
.

Another independent network, the
third

one,
in
c
ludes
several transcription factors,
and
gene
s
involved in
the
cell cycle
and in

Wnt, Notch
and
Jak
-
STAT

signaling pathways.

Mining the 623 genes of KUROV with the STRING database, searching for interactions with target genes, 324
genes with double contacts were connected to 1653 target genes and 377

genes were related to 3277 genes via a
single connection

(data not shown)
.
Taking one example of what this kind of mining can offer when using
Cytoscape for displaying results, we focused on two genes
from
the
1e sub
-
network

and
the
third

work:
Cep57
,
and
Lmo4
.

The
Cep57

gene

encodes a centrosomal
protein
, translokin,
which
localizes to the centrosome and has
a function in microtubular stabilization
. M
utation of this gene
in humans

generates the mosaic variegated
aneuploidy syndrome2

(Snape et al., 2011)
.
In our network analysis

this gene
was connected directly with 5

genes

(Fig. 4B),
three of them belong to the tubulin family. Mutation of
T
uba1a

in human
s

generate
s

lissencephaly3

characterized
by
microcephaly, pachygyria, an abnormally shaped corpus callosum, and
hypoplasia of the cerebellar vermis and brainstem

(Keays et al., 2007; Sohal et al., 2012)
.
From th
e STRING
database
Cep57

is i
ndirectly
connected
with any genes
th
rough a pool of 92

target genes (Fig.
4B
).

The general expression of these genes may be useful to predict their importance during development for the five
sensory organs

and might presents
specific pathology, see the section of

diseases related to genes found in
KUROV
.

One example like

Lmo4

(LIM domain only 4) from the
third

network is related directly to 7 other
genes present in KUROV and indirectly via 148 single target genes found in the
STRING databas
e

(Fig.
4C
).
Amongst these indirect target genes they are 5 Activating Transcription Factors (
Atf
) genes, 5 homeobox (
Hoxa
)
11



genes, 9 paired
-
box (
Pax
) genes

and 5

Pou domain transcription factors and several other transcription factors.
Lmo4

belongs to a family of transcriptional regulators that comprises two zinc
-
binding LIM domains. From our
data,
Lmo4

transcripts are strongly expressed in the inner ear mesenchyme, the semi
-
circular canals, and the
vibrissae follicles, and at a lower level
in vestibular sensory regions, the sensory retina and the olfactory organ.
These expressions correspond in part to previous observations reported for several tissues
(Kenny et al., 1998;
Sum et al., 2005)
. Targeted disruption of
Lmo4
, which is lethal at birth, results in a dysmorphogenesis of the
vesti
bular organs

with an absence of the three semicircular canals
(Deng et al., 2010)
. Furthermore, Cre
-
mediated disruption of
Lmo4

in the retina affects neuronal differentiation, leading to visual deficits
(Duquette et
al., 2010)
. It has also been shown that
Lmo4

is involved in development of the central nervous system, for
ins
tance its deletion affects neural tube closure and patterning of the thalamocortical connections
(Hahm et al.,
2004; Kashani et al., 2006)
. From these observations and our ISH data one may predict that
Lmo4

may be also
important for the development of vibrissae follicles, and possibly to all hair follicles. Moreover, it may be
essential for cochlear function as well to the olfactory organ.


Gene function
s

in relation to
Gene Ontology (
GO
)

annotation

Comparing the gene function

thanks to

GO annotation
s

for genes from the initial 2000 genes studied and the 623
of KUROV, we chose only five groups of genes that
generate
an important number of corresponding proteins
(Fig. 5
A). The most important group is the Protein binding followed by the Ion binding. Comparing the
distribution of gene products between the pool of 2000 genes and those of KUROV, the proportion of gene
prod
ucts increases for the Protein binding, Nucleic acid binding and Nucleotide binding in the KUROV pool of
genes compared to the overall 2000 genes. Looking in more detail in the distribution of t
he Protein binding
group (Fig. 5
A), Enzyme binding represents
the most important group followed by Receptor binding. Others
proteins bindings such as Transcription factor binding, Protein C
-
terminus domain specific, Heat shock,
Unfolded protein and Glycoprotein bindings are more represented in KUROV than in the 2000,

while the
R
eceptor bindings is less represented. Comparing some protein activities from the GO
annotation

between
KUROV and the 2000, the main difference is the lower number of gene products related to channel activities,
such as Ion channel activity and
G protein
-
c
oupled receptor activity (Fig. 5
B).

The KUROV pool of genes
presents a larger ratio of protein binding, nucleic acid binding and nucleotide binding proteins compared to the
initial 2000 genes.
This difference may be explained by the various cell

metabolic processes operating in the five
sensory organs at this period of development as these as been observed in the retina (Zhang et al., 2006), while in
the inner ear a switch seem to operate between genes involved in cell proliferation toward specif
ic cell
maturation
(
Sajan et al., 2007
).

Moreover, the lower number of ion channel activity and G protein
-
coupled
receptor proteins
reflects the same trend
and may be related to the fact that these sensory receptors have not yet
reached th
eir

functional statue
(
Romand et al., 1987
; Blackshaw et al., 2004;
Tian & Ma 2008
).


Network analysis of genes expressed in the olfactory organ

12



We performed additional network analysis focusing on one sensory organ, the olfactory organ, as an example.
The
number of genes outside KUROV (i.e. not expressed in the 5 sensory organs) and scoring positive for the
olfactory organ reached 552. From the STRING database along with Cytoscape analysis, 61 of these genes
presented an interconnection with themselves and
generated 8 networks with 2 major networks and 6 minor ones
(Fig. 6). The largest network (network 1) can be divided into two sub
-
networks, the first one containing

Socs4,

Il4r, Cblc, Synj2, Pip4k2c

and
Pik3cd
mainly involved in cytokine, Jak
-
STAT, Insulin

signaling and
I
nositol
phosphate pathways. The genes from the second sub
-
network are involved in axon guidance (
Srgap2, Nrp1)
,
focal adhesion (
Ptk2, Sos1, Shc
), or actin cytoskeleton (
Tiam1, Sos1, Fgf3
). In network 3 several genes are
involved in cell div
ision (
Seh1l, Spc24,
Cenpl
)
, while the fourth network is made of several s
p
icing factors and
transcription factors (
Sf3b5, Sf3a2, Nfx1
). The fifth network is mainly composed of voltage
-
dependent calcium
channels, and one gene involved in synaptic vesicle c
ycle (
Stx3
). The sixth network includes transcription
factors, the seventh one is composed of genes mainly involved in the MAPK signaling pathway (
Mapk14,
Dusp1, Dusp14
), while the last one includes only ribosomal proteins.
Out of 7 olfactory receptors tes
ted only 3
were expressed in this organ (
Olfr555, Olfr569
&

Olfr 592
), where

Olfr555
&
Olf568

present mutual
connections (data not shown) and are related to 155 target genes which are for a large majority olfactory
receptors
, (data not shown)
.


Gene
expression at the onset of olfactory function

Next, we compared the gene products obtained from the GO annotation between the 1175 transcripts expressed
in the olfactory organ (
Table S2 for a complete list
) and the 825 that are not expressed in this organ.

It is
interesting to note that gene products related to cell metabolism such as ligase activity and cytoskeletal protein
binding are more represented in the pool of genes expressed in the olfactory organ (Fig. 7). On the other hand,
gene products involved

in receptor activity such as G
-
protein coupled receptors (GPCRs) and ion channel
proteins are in lower number in the olfactory organ. The fact that the GPCRs are underrepresented may be
explained by the fact the olfactory function, i.e. detection of volat
ile odor molecules by G
-
protein coupled
odorant receptors, is not yet functioning at this stage of development
(Malnic et al., 2010)
. These observations
are consistent with data gained from the 5 sensory organs (KUROV) where proteins involved in channel
activities were underrepresented (Fig. 5), suggesting a lack of function for several sensory receptors duri
ng
in
utero

life. In mice the olfactory epithelium becomes organized into its mature pattern and neurogenesis is
initiated at around E13.5
-
E14.5
(Smart, 1971).

The overall number of mitotic figures decreases and becomes
localized primarily to the basal compartment of the olfactory epithelium where progenitor cells generate new
cells during later stages
(Beites et al., 2005)
.
By day E16 in the rat, the el
ectro
-
olfactogram is well developed,
and the receptor neurons are functional in that they respond to the vapors of odorous substances, although they
are not selective
(Gesteland et al., 1982)
. This observation suggests that in a few days maturation of the olfactory
epithelium proceeds very

rapidly in order to be functional around birth. Although still relatively immature the
olfactory function undergoes significant development postnatally, developing very early a preference for familiar
odors which is very important for pups survival
(Levy et al., 2004)
.

Many factors, including diseases,
influence the normal

ability to smell,
(
see Do
ty
,

2009 for a review
)

and its dysfunction is now

13



known to be among the earliest ‘‘preclinical’’ signs of Alzheimer’s disease and sporadic Parkinson’s
disease
, (Devanand et al 2008; Ross et al., 2008).


Human
known
diseases

related to genes
found in KUROV

Using the Morbid database, we search
ed

for genes
in

the KUROV list that could be related to known human
diseases. From the 623 genes, 70 were related to diseases with 15 related to specific syndromes (Table 5), such
as:
Aldh3a2

for
Sjogren
-
Larsson syndrome
,
Hsp4

for Hermansky
-
Pudlak syndrome 4,
Slc12a1

for Bartter
syndrome (type 1) and
Crebbp

for Rubinstein
-
Taybi syndrome. Th
e

lat
t
er syndrome is characterized by mental
retardation, hearing loss, hirsutism, glaucoma and nose malformations
(Roelfsema and Peters, 2007)
.

Crebbp

(encoding a CREB
-
binding protein)
was found to be weakly expressed in
the five

sensory organs (Fig
s
. 8
A
-
E)
and presented several interrelations with genes from KUROV and f
rom the STRING database (Fig
s
. 8
F
-
H). This
gene is directly connected to 11 genes present in KUROV, and indirectly via 169

genes from the STRING
database exhibitin
g interactions with two initial genes

and

187

with single target genes
. These complex
interactions

likely relate to functional processes

affect
ing

the development or the function of sensory organs.
This is not unexpected as this gene is involved in several basic cellular pathways operating during development
such as Wnt, Notch, TGF
-

, Jak
-
STAT signaling, adherens junctions, and cell cycle
,

and

may b
e involved

in
specific diseases like
DiGeorge

syndrome
,
ovarian tumorigenesis
and prostate cancer
(
Wurdack et al., 2005,
Boyer et al., 2010
,
Liu et al., 2012
)
and cognitive function like long
-
term memory formation, (
Maguschak
&


2010
).

Another interesting example is coming from
Hps4
,
which
exhibits
a weak expression in all
five
sensory organs.
M
utation
of its human homologue
is responsible for Hermansky
-
Pudlak syndrome type 4 (Table 5),
an
autosomal recessive disorder
with

oculocutaneous albinism and bleeding attributable to storage
-
pool

deficien
cy

in lysosome
-
related organelles such as melanosomes and platelet
-
dense granules.

This syndrome is characterized
by pulmonary fibrosis, variable hair and skin hypopigmentation
, and

lower visual acuity

with macular
translucency (
Anderson et al., 2003).
From the STRING database we found that this gene is connected to 3
Hps

genes
,
Hps1
,

Hps3, Hps6
,
whose mutation
s

in human
result in

similar pathologies
(Gahl, 2012)
.
The rodent
model for this syndrome is the
light ear

mouse that is deficient in HPS4 and HPS1 proteins
.

Th
ese

mutation
s

alters

the formation of melanosomes in melanocytes

(Suzuki et al., 2002)
, affecting the coat color
; it

could also
be related to inner ear function since melanocytes are important for the function of this sensory organ

(Cable et
al., 1994)
.
F
u
rther

work may demonstrate

that mutation
s

affecting

Hps4
could affect in one way or another, all

five

sensory organs.




14



C
ONCLUSION

We report the ISH express
i
on of 2000 transcripts in 25 tissues
of E14.5 m
ice

as

this

can be observed on more
than 11

000 pictures
displayed
on
a

WWW

database
.
We believe that your 2000 genes correspond to around a
10% random sample of the whole genome of the mouse, (Waterston et al., 2002; Carninci et al., 2007).
Therefore, the number of genes found for each organ may represent approximately 8
-
10% of the whole g
enome,
this would suggest for example that in the olfactory organ approximately 11

000 genes from the whole genome
could be involved at this stage of development. However, this conclusion has to be made with caution because
this hypothesis may not apply fo
r a specific group of genes, for instance we tested only 7 olfactory receptors
while there are more than 1000 of such related genes in the mouse olfactory epithelium (Zhang & Firestein,
2009).
Bioinformatic a
nalysis by hierarchical tissue clustering

shows
that grouping of different tissues
corresponds either to a common embryological origin like mesenchyme derived tissues or their prospective
function as for nervous tissues. Another interesting observation is that our sampling of 2000 genes mimics a
functio
nal aspect or an embryological origin that reinforces our random sampling of our 2000 genes.
Data
mining has been restricted to five sensory organs i.e. the Kölliker’s organ for the cochlea, the utricule for the
balance organ, the sensory retina

for the ey
e
, the olfactory organ

for the nose

and the vibrissae follicles

from the
muzzle

grouped in a pool termed KUROV
. Gene expressions were
often very restricted
and

have never been
described before
. An
important percentage of

probes
tested by ISH (around 30%)
w
ere

expressed all together in
the five sensory
epithelia
. However, for a specific sensory organ more than half of gene
s

tested were found to be
present, except for
the sensory receptors of the inner ear

where the number was lower
.
Ninety five percent of
transcript
expressions were in the weak or medium range

that could correspond to what has been found in gene
expression profiles in normal cells
, (Zhang et al., 1997)
.

Further data mining
with STRING database and
Cytoscape

on gene expressions
from
KUROV
and the olfactory organ
have revealed
several
complex networks
of gene pathways involved in
cell cycle, apoptosis, cytoskeleton regulation, neural development and
transcription where Wnt, Notch, Jak
-
STAT, and Hedgehog signaling pathways play important
meta
bolic
processes operating at this stage of development
.
Comparing gene products
with GO annotation
between our
initial data base (2000 probes)
and those from KUROV or the olfactory organ
show that the number of gene

products

involved in
the function of
sen
sory cells

are less numerous suggesting that
these
sensory receptors have
not yet
reached

a functional statue.

Using the Morbid database, we searched for
the 623
genes
from

the KUROV
list that could be related to known human diseases. From the
se

genes

with

specific gene expression in the five
sensory organs
, around 10% were related to diseases
, while
some

of them were related to sensory
function
impairments. It could be possible to predict in the future that on the basis of their transcript expression
s

in the
five sensory organs more and more
impairments

will be discovered affecting the
various
sensory functions.





15



ACKNOWLEDGMENTS

This work was supported
by funds from CNRS, INSERM, Université de Strasbourg,
and grants from
Agence Nationale de la
Recherche (ANR)

and

Fondation pour la Recherche Médicale (FRM
)
.

REFERENCES

Alvarez
-
Bolado, G., and Eichele, G. (2006). Analysing the developing brain transcriptome with the
GenePaint platform. J Physiol 575, 347
-
352.

Anderson PD
,
Huizing M
,
Claassen DA
,
White J
,
Gahl WA
. 2003
Hermansky
-
Pudlak

syndrom
e

type

4
(HPS
-
4): clinical and molecular characteristics.
Hum Genet.

113:10
-
17.

Anderson WJ, Zhou Q, Alcalde V, Kaneko OF, Blank LJ, Sherwood RI, Guseh JS, Rajagopal J, Melto
n
DA. 2008
Genetic targeting of the endoderm with
claudin
-
6CreER.

Dev Dyn
. 237, 504
-
512.

Back SH
,
Scheuner D
,
Han J
,
Song B
,
Ribick M
,
Wang J
,
Gildersleeve RD
,
Pennathur S
,
Kaufman
RJ
.
2009 Translation attenuation through eIF2alpha phosphorylation prevents oxidative stress and
maintains the differentiated state in beta cells.
Cell Metab.

10, 13
-
26.


Beites, C.L., Kawauchi, S., Crocker, C.E., and Calof, A.L. (2005). Identification and molecular regulation
of neural stem cells in the olfactory epithelium. Exp Cell Res 306, 309
-
316.

Blackshaw S, Harpavat S, Trimarchi J, Cai L, Huang H, Kuo WP, Weber G, L
ee K, Fraioli RE, Cho SH,
Yung R, Asch E, Ohno
-
Machado L, Wong WH,
Cepko CL
. 2004
Genomic analysis of mouse retinal
development.

PLoS Biol
. E247.

Boyer A
,
Goff AK
,
Boerboom D
. 2009
WNT

signaling in

ovarian follicle biology and tumorigenesis.
Trends Endocrinol Metab.
21, 25
-
32.

Breer, H., Fleischer, J., and Strotmann, J. (2006). The sense of
smell: multiple olfactory subsystems.
Cell Mol Life Sci 63, 1465
-
1475.

Cable, J., Huszar, D., Jaenisch, R., and Steel, K.P. (1994). Effects of mutations at the W locus (c
-
kit) on
inner ear pigmentation and function in the mouse. Pigment Cell Res 7, 17
-
32.

Carninci P, Hayashizaki Y 2007 "Noncoding RNA transcription beyond annotated genes".
Curr. Opin.
Genet. Dev.

17
, 139

144.

Carson JP
,
Thaller C
,
Eichele G
. 2002 A transcriptome atlas of the mouse brain at cellular resolu
tion.
Curr Opin
Neurobiol.

12, 562
-
265.

Couloigner

V
,
Teixeira M
,
Sterkers O
,
Rask
-
Andersen H
,
Ferrary E
.
2004 Le sac endolymphatique: ses
fonctions au sein d
e l’oreille interne.
Med Sci (Paris).

20, 304
-
310.

Cristobal R
,
Wackym PA
,
Cioffi JA
,
Erbe CB
,
Roche JP
,
Popper P
.
2005 Assessment of differential gene
expression in vestibular epithelial cell types using microarray analysis.
Brain Res Mol Brain Res.

133,19
-
36.

16



Deng, M., Pan, L., Xie, X., and Gan, L. (2010). Requirement for Lmo4 in the vestibular morphogenesis
of mouse inner ear. Dev Biol

338
, 38
-
49.

Depew, M.J., Lufkin, T., and Rubenstein, J.L. (2002). Speci
fication of jaw subdivisions by Dlx genes.
Science

298
, 381
-
385.

Devanand DP
,
Liu X
,
Tabert MH
,
Pradhaban G
,
Cuasay K
,
Bell K
,
de Leon MJ
,
Doty RL
,
Stern Y
,
Pelton GH
. 2
008 Combining early markers strongly predicts conversion from mild cognitive
impairment to Alzheimer's disease.
Biol Psychiatry.

64, 871
-
879.

Diez
-
Roux, G., Banfi, S.,
Sultan, M., Geffers, L., Anand, S., Rozado, D., Magen, A., Canidio, E., Pagani,
M., Peluso, I.
, et al.

(2011). A high
-
resolution anatomical atlas of the transcriptome in the mouse
embryo. PLoS Biol

9
, e1000582.

Duquette, P.M., Zhou, X., Yap, N.L.,
MacLaren, E.J., Lu, J.J., Wallace, V.A., and Chen, H.H. (2010). Loss
of LMO4 in the retina leads to reduction of GABAergic amacrine cells and functional deficits. PLoS
One

5
, e13232.

Emerich, D.F., Skinner, S.J., Borlongan, C.V., Vasconcellos, A.V., and Th
anos, C.G. (2005). The choroid
plexus in the rise, fall and repair of the brain. Bioessays

27
, 262
-
274.

Gahl, W.A. (2012). Hermansky
-
Pudlak Syndrome. In: Pagon RA, Bird TD, Dolan CR, Stephens K, editors
GeneReviews Seattle (WA): University of Washington,
Seattle.

Gesteland, R.C., Yancey, R.A., and Farbman, A.I. (1982). Development of olfactory receptor neuron
selectivity in the rat fetus.
Neuroscience

7
, 3127
-
3136.

Gofflot, F., Chartoire, N., Vasseur, L., Heikkinen, S., Dembele, D., Le Merrer, J., and Auwe
rx, J. (2007).
Systematic gene expression mapping clusters nuclear receptors according to their function in the
brain. Cell

131
, 405
-
418.

Gorry P
,
Lufkin T
,
Dierich A
,
Rochette
-
Egly C
,
Decimo D
,
Dolle P
,
Mark M
,
Durand B
,
Chambon P
.
1994
.
The cellular
retinoic acid binding protein I is dispensable
.
Proc Natl Acad Sci U S A

91
,
9032

9036
.

Hahm, K., Sum, E.Y., Fujiwara, Y., Lindeman, G.J., Visvader, J.E., and Orkin, S.H. (2004). Defective
neural tube closure and anteroposterior patterning in mice lacking
the LIM protein LMO4 or its
interacting partner Deaf
-
1. Mol Cell Biol

24
, 2074
-
2082
.

Jave
-
Suarez LF, Winter H, Langbein L, Rogers MA, Schweizer J. HOXC13 is involved in the regulation of
human hair keratin gene expression. J Biol Chem. 2002
,
277:3718

37
26

Kashani, A.H., Qiu, Z., Jurata, L., Lee, S.K., Pfaff, S., Goebbels, S., Nave, K.A., and Ghosh, A. (2006).
Calcium activation of the LMO4 transcription complex and its role in the patterning of
thalamocortical connections. J Neurosci

26
, 8398
-
8408.

Keays, D
.A., Tian, G., Poirier, K., Huang, G.J., Siebold, C., Cleak, J., Oliver, P.L., Fray, M., Harvey, R.J.,
Molnar, Z.
, et al.

(2007). Mutations in alpha
-
tubulin cause abnormal neuronal migration in mice and
lissencephaly in humans. Cell

128
, 45
-
57.

17



Kenny, D.A.
, Jurata, L.W., Saga, Y., and Gill, G.N. (1998). Identification and characterization of LMO4,
an LMO gene with a novel pattern of expression during embryogenesis. Proc Natl Acad Sci U S A

95
,
11257
-
11262.

Kessel, R.G., and Kardon, R.H. (1979). Tissues and
Organs: a text
-
atlas of scanning electron
microscopy. MH Freeman & Co, San Francisco.

Kim SJ
,
Dix DJ
,
Thompson KE
,
Murrell RN
,
Schmid JE
,
Gallagher JE
,
Rockett JC
. 2006 Gene
expression in head
hair

follicles plucked from men and women.
Ann Clin Lab Sci.
36, 115
-
126.

Kitsios GD
,
Tangri N
,
Castaldi PJ
,
Ioannidis JP
. 2010 Laboratory
mouse

models for the hum
an
genome
-
wide associations. PLoS One 5, e13782.

Lau, C., Ng, L., Thompson, C., Pathak, S., Kuan, L., Jones, A., and Hawrylycz, M. (2008). Exploration
and visualization of gene expression with neuroanatomy in the adult mouse brain. BMC
Bioinformatics

9
,
153.

Lander ES, Linton LM, Birren B, Nusbaum C, Zody MC, et al. 2001 Initial sequencing and analysis of
the human genome. Nature, 409,

860
-
921.

Lefkowitz, G.K., and Gleeson, J.G. (2009). Aurora A moonlights in neurite extension. Nat Cell Biol

11
,
1053
-
1054
.

Levy, F., Keller, M., and Poindron, P. (2004). Olfactory regulation of maternal behavior in mammals.
Horm Behav

46
, 284
-
302.

Liu X
,
Li CH
,
Huang G
,
Ding C
,
Liu H
. 2012 Correlation analysis of
JAK
-
STAT

pathway components on
prognosis of patients with prostate cancer.
Pathol Oncol Res.

18
,
17
-
23.

Magdaleno, S., Jensen, P., Brumwell, C.L., Seal, A., Lehman, K., Asbury, A., Cheung, T., Cornelius, T.,
Batten, D.M., Eden, C.
, et al.

(2006). BGEM: an in situ hybridization database of gene e
xpression in
the embryonic and adult mouse nervous system. PLoS Biol

4
, e86.

Mallo, M. (2003). Formation of the outer and middle ear, molecular mechanisms. In: Development of
Auditory and Vestibular Systems
-
3 Molecular Development of the Inner Ear Eds R Ro
mand & I Varela
-
Nieto

Elsevier/Academic Press, pp. 85
-
113.

Malnic, B., Gonzalez
-
Kristeller, D.C., and Gutiyama, L.M. (2010). Odorant Receptors. In: Menini A, Ed.
The Neurobiology of Olfaction. CRC Press, Boca Raton (Fl).

Maguschak KA
,
Ressler KJ
. 2010
Wnt

signaling

in amygdala
-
dependent learning and memory.
J
Neurosci.

31
,
13057
-
13067.

Matsubara A
,
Miyashita T
,
Mori T
,
Akiyama K
,
Inamoto R
,
Mori N
. 2011 The mRNA o
f claudins is
expressed in the endolymphatic sac epithelia.
Auris Nasus Larynx.

Aug 24. [Epub ahead of print]

Minowa

O, Ikeda K, Sugitani Y, Oshima T, Nakai S, Katori Y, Suzuki M, Furukawa M, Kawase T, Zheng
Y, Ogura M, Asada Y, Watanabe K, Yamanaka H, Gotoh S, Nishi
-
Takeshima M, Sugimoto T, Kikuchi T,
Takasaka T, Noda T. 1999
Altered cochlear fibrocytes in a mouse model of DFN3 nonsyndromic
deafness.

Science
. 285,

1408
-
1411.

18



Moon SK, Woo JI, Lee HY, Park R, Shimada J, Pan H, Gellibolian R, Lim DJ. 2007
Toll
-
like receptor 2
-
dependent NF
-
kappaB activation is involved in nontypeable Haemophilus influenzae
-
induced
monocyte chemotactic protein 1 up
-
regulation in the spiral ligament fibrocytes of the
inner ear
.

Infect Immun
. 75:3361
-
3372.

Neidhardt, L., Gasc
a, S., Wertz, K., Obermayr, F., Worpenberg, S., Lehrach, H., and Herrmann, B.G.
(2000). Large
-
scale screen for genes controlling mammalian embryogenesis, using high
-
throughput
gene expression analysis in mouse embryos. Mech Dev

98
, 77
-
94.

Ng, L., Bernard,
A., Lau, C., Overly, C.C., Dong, H.W., Kuan, C., Pathak, S., Sunkin, S.M., Dang, C.,
Bohland, J.W.
, et al.

(2009). An anatomic gene expression atlas of the adult mouse brain. Nat
Neurosci

12
, 356
-
362.

Nguyen D
,
Xu T
. 2008 The expanding role of
mouse

genetics

for understanding
human

biology and
disease.
Dis Model Mech.

1, 56
-
66.

Potter CS, Pruett ND, Kern MJ, Baybo MA, Godwin AR, Potter KA, Peterson RL, Sundberg JP,
Awgulewitsch A. 2011
The nude

mutant gene Foxn1 is a
HOXC13

regulatory target during hair follicle
and nail differentiation.

J Invest Dermatol
. 131, 828
-
837.
Roelfsema, J.H., and Peters, D.J. (2007).
Rubinstein
-
Taybi syndrome: clinical and molecular overview. Expert Rev Mol Med

9
,
1
-
16.

Romand R, Despres G, Giry N. 1987
Factors affecting the
onset

of
inner ear

function
.

Hear Res
. 28, 1
-
7.

Ross GW, Petrovitch H, Abbott RD, et al. Association of olfactory dysfunction with risk

for future
Parkinson’s disease. Ann Neurol 63, 167

173

Rossant, J., and Tam, P.P. (2002). Mouse development. Patterning, Morphogenesis, and
Organogenesis. Academic Press, San Diego.

Ruberte E
,
Friederich V
,
Morriss
-
Kay G
,
Chambon P
.
1992
.
Differential
distribution patterns of
CRABP I and CRABP II transcripts during mouse embryogenesis
.
Development

11
5,
973

987
.

Sajan, S.A., Warchol, M.E., and Lovett, M. (2007). Toward a systems biology of mouse inner ear
organogenesis: gene expression pathways, patterns

and network analysis. Genetics

177
, 631
-
653.

Smart

IH. 1971
Location and orientation of mitotic figures in the developing mouse
olfactory

epithelium.

J Anat
. 109, 243
-
251.

Snape, K., Hanks, S.,
Ruark, E., Barros
-
Nunez, P., Elliott, A., Murray, A., Lane, A.H., Shannon, N.,
Callier, P., Chitayat, D.
, et al.

(2011). Mutations in CEP57 cause mosaic variegated aneuploidy
syndrome. Nat Genet

43
, 527
-
529.

Sohal, A.P., Montgomery, T., Mitra, D., and Rame
sh, V. (2012). TUBA1A Mutation
-
Associated
Lissencephaly: Case Report and Review of the Literature. Pediatr Neurol

46
, 127
-
131.

Stansberg

C, Vik
-
Mo AO, Holdhus R, Breilid H, Srebro B, Petersen K, Jørgensen HA, Jonassen I, Steen
VM. 2007
Gene

expression profiles in rat brain disclose CNS signature genes and regional patterns of
functional specialisation.

BMC Genomics
. Apr 4;8:94

19



Steel, K.P., and Barkway, C. (1989). Another role for
melanocytes: their importance for normal stria
vascularis development in the mammalian inner ear. Development

107
, 453
-
463.

Stevens, C.A. (2009). Rubinstein
-
Taybi Syndrome. In: Pagon RA, Bird TC, Dolan CR, Stephens K, editors
GeneReviews [Internet] Seattle

(WA): University of Washington, Seattle.

Sum, E.Y., O'Reilly, L.A., Jonas, N., Lindeman, G.J., and Visvader, J.E. (2005). The LIM domain protein
Lmo4 is highly expressed in proliferating mouse epithelial tissues. J Histochem Cytochem

53
, 475
-
486.

Suzuki,
T., Li, W., Zhang, Q., Karim, A., Novak, E.K., Sviderskaya, E.V., Hill, S.P., Bennett, D.C., Levin,
A.V., Nieuwenhuis, H.K.
, et al.

(2002). Hermansky
-
Pudlak syndrome is caused by mutations in HPS4,
the human homolog of the mouse light
-
ear gene. Nat Genet

3
0
, 321
-
324.

Swetloff, A., and Ferretti, P. (2005). Changes in E2F5 intracellular localization in mouse and human
choroid plexus epithelium with development. Int J Dev Biol

49
, 859
-
865.

Thalmann R
,
Ignatova E
,
Kachar B
,
Ornitz DM
,
Thalmann I

2001 Development and maintenance of
otoconia: biochemical consi
derations.
Ann N Y Acad Sci.

942, 162
-
178.

Tian H, Ma M. 2008
Differential development
of odorant receptor expression patterns in the
olfactory

epithelium: a quantitative analysis in the mouse septal
organ
.

Dev Neurobiol
. 68,
476
-
86.

Uehara S
,
Izumi Y
,
Kubo Y
,
Wang CC
,
Mineta K
,
Ikeo K
,
Gojobori T
,
Tachibana M
,
Kikuchi T
,
Kobayashi T
,
Shibahara S
,
Taya C
,
Yonekawa H
,
Shiroishi T
,
Yamamoto H
. 2009 Specific expression
of
Gsta4

in mouse cochlear melanocytes: a novel role for hearing and melanocyte differentiation.
Pigment Cell Melanoma Res.

22, 111
-
119.

Visel A
,
Thaller C
,
Eichele G
. 2004 GenePaint.org: an atlas of gene expression patterns in the mouse embryo.
Nucleic Acids Res.
, 32, D552
-
556.

Visel, A., Carson, J., Oldekamp, J., Warnecke, M., Jakubcakova, V., Zhou, X., Shaw, C.A., Alvarez
-
Bolado, G., and
Eichele, G. (2007). Regulatory pathway analysis by high
-
throughput in situ hybridization.

PLoS Genet

3
, 1867
-
1883.

Waterston, RH
,

Lindblad
-
Toh K, Birney E, Rogers J, Abril JF, Agarwal P
,
Agarwala R, Ainscough R.
Alexandersson M, An P. et al.
2002 Initial

sequencing and comparative analysis of the mouse
genome.
Nature,

420
, 520
-
562.

Wurdak H,
Ittner LM, Lang KS. Leveen P, Suter U, Fischer JA, Karlsson S, Born W, Sommer L. 2005

Inactivation of TGF
-
beta signaling in neural crest stem cells leads to multiple defects reminiscent of
DiGeorge syndrome.

Genes Dev. 19: 530
-
535,

Zhang L
,
Zhou W
,
Velculescu VE
,
Kern SE
,
Hruban RH
,
Hamilton SR
,
Vogelstein B
,
Kinzler KW
.
1997 Gene expression profiles in normal and cancer cells. Science 276
,
1268
-
1272.

Zhang SS
,
Xu X
,
Liu MG
,
Zhao H
,
Soares MB
,
Barnstable CJ
,
Fu XY
. 2006 A
biphasic

pattern of
gene expression during mouse retina development.
BMC Dev Biol.

6,

48.

Zhang, X., and Firestein, S. (2009). Genomics of olfactory receptors. Results Probl Cell Differ

47
, 25
-
36.


20



FIGURE LEGENDS

Figure 1
. Tissue dendrogram, based upon a hierarchical clustering using average linkage of the
differentially expressed genes. Three main nodes with several sub
-
nodes can be distinguished. Node
A is related to genes from tissues mainly derived from sensory regions

of the inner ear. Node B is
more heterogeneous, and includes various mesenchyme
-
derived tissues such as different cartilages
or bones, and secretory organs such as the stria vascularis. This node also includes the inner ear
endolymphatic organ and the hin
dbrain choroid plexus, as well as mesenchymal tissues from the
inner ear and middle ear. Node C mainly relates to genes expressed in nervous tissues such as the
retina, and ectodermal/mesenchymal tissues from the middle ear and the vibrissae follicles.

T
correspond to the tissue’s number found
in
the Gene annotation da
t
abase.

Expliquer la barre 5e+01

Figure 2
.
Examples of gene expression patterns in the sensory organs.

A & B: Two expressed genes in the basal cochlear canal. The cochlear canal is delineated

by dashed
lines: Ko: Kölliker’s organ presents in the ventral region, Iss: inner spiral sulcus may includes the
prospective Reissner’s membrane and the outer spiral sulcus (Oss), Oc: otic capsule, sagital section.

The patchy expression of
Ctgf

(Connective tissue growth factor) seems to be restricted to the
Kölliker’s organ extending toward the outer spiral sulcus (Oss) and the otic capsule (Oc). Interesting
enough in the Kölliker‘s organ a region without expression (arrow) separates what could
be the greater
epithelial ridge from the lesser epithelia ridge. The transcript expression of
Shc3

(Src homology 2
domain
-
containing transforming protein C3) is visible in the basal canal of the cochlea and restricted
to a small region of the Kölliker’s or
gan (arrows).

Scale bar : 50
µm
.

C & D: Expression of two genes in the utricule from the vestibular part of the inner ear. The
expression of the
Cd9

(Cd 9 antigen)

is visible in the prospective sensory region (Sr) of the utricule as
well as the non
-
sensory

region (Nsr), (large arrow). The two horizontal arrows points toward the
separation between the sensory region and the non
-
sensory region. Sagital sections.

Scale bar: 100
µm
.

E & F: Two examples of transcripts expressed in the retina:
Mid1

(Midline 1) and

Fubp1

(Far upstream
element (FUSE) binding protein 1).

Rpe: the retinal pigmented epithelium. Scale bar: 0.5 mm.

G & H: Two samples of expressed gene as observed on sagital sections of the olfactory organs.

The
gene
Gp38
(
Glycoprotein 38 or podoplanin) is observed in the olfactory epithelium (Oe) and the
cartilage primordia of turbinate bones (Ct).
The probe for

Igsf4a

(
Immunoglobulin superfamily, member
4A, transcript variant 2) is present all over the olfactory organ incl
uding the respiratory epithelium
(Re). Scale bar: 200 µm.

I: Sagital sections on the primordia of
vibrissae

follicle
s
. Depending of the level of the section in the
upper lip, different regions of the
vibrissae

present a mRNA expression. The
Mif

gene (Macro
phage
migration inhibitory factor) is expressed in several regions of the
vibrissae

follicles including the
follicle cells. Expression is also observed outside the
vibrissae

in the surrounding mesenchyme (Me).
Scale bar: 0.5 mm.




21



Figure 3
.
A: Percentages

of expressed genes from 2000 genes analyzed in the

EURExpress in situ
hybridization data collection. Percentages are given relative to the level of labeling in the sensory
region of five sensory organs: the retina, the olfactory organ, the vibrissae folli
cles,
the
Kölliker’s
organ in the basal cochlear canal, and the utricular macula from the vestibular organ.

B:
D
iagrams
illustrating the pools of genes expressed in common between two different sensory organs. The
number of genes expressed in Kölliker’s or
gan and the utricle is much lower compared to the three
other sensory organs.

Enlever le Not applicable

à voir


Figure 4.
Network analysis of genes expressed in the five sensory organs.

Genes that were found to be
expressed in the five sensory organs were
grouped in a pool termed “KUROV”, where K stands for the
Kölliker's organ, U for the sensory region of the utricle, R for the sensory retina, O for the olfactory
organ and V for the vibrissae

follicles
. For this analysis we have used

the STRING database al
ong
with Cytoscape in order to find the possible interactions between the 623 genes that are common to the
five sensory tissues.
A
: Out of 623 genes, only 163

presented direct interactions where they could be
grouped into 7 major networks, although other l
ess important networks were also found. The first
network is mainly composed of ribosomal proteins and mitochondrial ribosomal proteins showing
numerous interactions. This network is linked through
Eif2s1

(Translation initiation factor2, subunit
1alpha) to a second network, whose genes are involved in protein processing. The first network is also
connected to a third one trough
Rps27l

(Ribosomal protein S27 like), where genes are linked to several
cellular

processes such as focal adhesion, cytoskeleton regulation and neural development. This
network is related to a fourth one through
Aurka

(Aurora kinase A) where

several proteosome genes

are present
.
Csnk1a1
involved in Wnt signaling, and Hedgehog signaling
pathways makes the
connection with the fifth network mainly made of genes involved with tubulin, actin, centrosomal and
cilia proteins. A sixth network independent from the previous ones is made of genes mainly involved
in transcription. Another (seventh)
independent network, includes several transcription factors.
B
:
Examples of mining trough the STRING database for two genes from the fifth

and
the seventh
networks:
Cep57
and
Lmo4
.
The
Cep57

gene encodes a centrosomal protein.
This gene i
s connected

directl
y with 5 genes from KUROV
, three of them belong
ing

to the tubulin family
, and

is indirectly
connected to a p
ool of 92 target genes
.

C
:
Lmo4

(LIM domain only 4) is
directly
related to 7 genes

present in KUROV, and
via a single connection to 148 outside target genes.








Figure 5.
Comparison
of

the distribution of gene functions using GO annotation between the initial
pool of 2000 genes and those from KUROV.
A
: Several important functions appear in the pool of
20
00 genes, the five most represented being Protein binding, Ion binding and three other groups with
22



approximately the same percentage: Nucleic acid binding, Nucleotide binding, and Hydrolase activity.
Out of the 2000 genes 359 were not found in the GO datab
ase, while 503 were not analyzed

included
in the
Others & no GO

section
.

The distribution of GO annotations for the 623 genes of KUROV show
a larger percentage of Proteins binding, as well as the next three groups of proteins. Looking in more
detail at the

distribution of Protein binding annotations, the percentage of Enzymes is more important
for KUROV, as well as for Transcription factor binding proteins, Proteins C
-
terminus, Heat shock,
Unfolded proteins and Glycoproteins, while the percentage of Recepto
rs is reduced.

B
: Comparison
between the main protein activities between the initial 2000 genes
(blue bars)
and those from KUROV

(red bars)
. This analysis shows that proteins
with

channel activity such as Ion channel activity and G
protein
-
coupled receptor

activity are in lower number in the KUROV pool of genes.


Figure 6.
Network analysis of g
enes outside KUROV, expressed in the olfactory sensory region. For
this analysis we have used

the STRING database in order to find possible interactions between 552
genes expressed in the olfactory organ

(but not belonging to the KUROV list), Cytoscape for
displaying the results along with the KEGG database for possible functional pathways. Out of 552
genes only 61presented direct interactions where the majority of th
em can be grouped into 2 major
networks and 6 minor ones. The first major network (1) can be divided into two sub
-
networks, with
genes involved in Jak
-
STAT signaling, insulin signaling and inositol phosphate signaling pathways.
The other sub
-
network includ
es genes involved in axon guidance, focal adhesion or actin cytoskeleton
and neurotrophin function. The second network (2) can also be divided into two sub
-
networks
involved in centrosomal proteins, in cellular processes and in Wnt and Tgf
-


signaling path
ways. The
second sub
-
network is mainly involved in cell cytoskeleton.
The 3
rd

network is composed of genes
involved in nuclear pore complex, centromere and kinetochore, while the 4
th

is involved in the
spliceosome and in nuclear transport. The 5
th

network
is mainly composed of calcium channel voltage
dependent and of synaptic vesicle gene like. The 6
th

network is composed of transcription factors,
while the last ones

are mainly made of genes involved in the MAPK signaling pathway, or encoding
ribosomal prot
eins.








Figure 7.
Comparison of the distribution of gene functions using GO annotation between genes
expressed or not e
xpressed in the olfactory organ.

To make possible the comparison between

the
two
sets of gene products, one
expressed in

the olfactory epithelium (1175 transcripts) and th
e other scored
negative

(825 transcripts), the number of genes w
as equalized to a

same reference number: 1000.
(
Pourquoi ne pas l'exprimer plus

simplement en pourcentages ? 40,

par exemple
,

veut donc dire
4 %
?
)

23



Gene products from two cellular functions were compared, one related to basic cellular function
involved during development such as the Ligase activity and Transcription cofactor activity, and one
in relation with receptor activity such as G
-
protein

coupled receptor activity and Ion channel activity.
Gene products involved in basic cellular functions during development such as Cytoskeletal protein
bindings were more
represented

in the pool of genes

expressed in

the olfactory organ
.

At the opposite,
g
ene products involved in receptor activity were less numerous

in this pool of genes
.


Figure 8.
Expression of
Crebbp

(a gene involved in Rubinstein
-
Taybi syndrome) in the five sensory
organs and its relationship with other genes.
A
-
E
: ISH
expression
patterns in

the basal cochlear canal
(A)

and the utric
le (B)
, both

delineated by dashed lines
,
The approximate sensory regions of these two
organs are delimited by red lines

(
préférable de les enlever
)
,

the retina

(C)
, the olfactory organ

(D)

and
the upper

lip

(E)

where sever
al

vibrissae follicles are

present.

F
:
STRING analysis showing
direct

interactions
of Crebbp

with 11 genes present in KUROV.

G
:
Crebbp

and the eleven
KUROV
genes
directly connected to

it show

interactions with two initial genes

with
169
other genes
from

the
STRING database.

H: diagram showing the single relation of
Crebbp

and the 11 genes from KUROV
with 187 genes as found from the STRING database.