FEBS Letters 581 (2007) 4884–4890

Early upregulation in nasal epithelium and strong expression inolfactory bulb glomeruli suggest a role for Aquaporin-4 in olfaction

Jan Gunnar Sørbø, Svein Erik Moe, Torgeir Holen*

Center for Molecular Biology and Neuroscience (CMBN), University of Oslo, Norway

Received 15 June 2007; revised 5 September 2007; accepted 10 September 2007

Available online 19 September 2007

Edited by Beat Imhof

Abstract Aquaporin-4 (AQP4) has been reported to be upreg-ulated post-partum in pregnancy and in early lung development.Several technical challenges exist in measuring AQP4 proteinlevels, among them sensitivity to detergent solubilization, sampleheating and gel composition. Here we have optimized quantifica-tion of AQP4 using immuno-blots. Using improved methodologywe find no evidence for AQP4 upregulation post-partum or in theearly lung development. However, in the nasal epithelium AQP4is upregulated as early as in the brain. Furthermore, AQP4 isstrongly expressed in the glomerulus, the synaptic unit of theolfactory bulb, suggesting a role for AQP4 in olfactory function.� 2007 Federation of European Biochemical Societies. Publishedby Elsevier B.V. All rights reserved.

Keywords: Aquaporins; AQP4; Membrane protein; Olfaction

1. Introduction

Aquaporins are a sequence-similarity family of 12 mamma-

lian water channels that facilitate water and glycerol transport

over cell membranes [1]. Whereas the role of some aquaporins

yet remains to be elucidated, aquaporin-facilitated water trans-

port is particularly well studied in the kidneys, where several

aquaporins are expressed: AQP1 in the proximal tubuli and

descending thin-limb, and AQP2, AQP3 and AQP4 in the col-

lecting ducts. The action of the anti-diuretic hormone arginine

vasopressin increases water reabsorption in the collecting duct

by recruiting AQP2 to the apical membrane, in response to in-

creased plasma osmolality and decreased effective circulating

volume, while AQP3 and AQP4 provide water transport over

the basolateral membrane [1–3].

AQP4 is expressed most strongly in the brain in rats [4,5],

humans [5,6] and mice [7]. A series of studies over the last 12

years have established that AQP4 is found at the interfaces be-

tween the brain and blood, and brain and cerebrospinal fluid,

respectively. AQP4 is enriched in astroglial endfeet facing

blood vessels, and in the ependymal cells lining the brain ven-

tricles [5,8,9]. These studies have led to the hypothesis that

AQP4 is involved in water homeostasis in the mammalian

brain [8]. Important clues about the role of AQP4 in the devel-

opment of brain water imbalance have been revealed by

knockout studies [10]. Strikingly, AQP4 knockout mice were

*Corresponding author.E-mail address: torgeir.holen@medisin.uio.no (T. Holen).

0014-5793/$32.00 � 2007 Federation of European Biochemical Societies. Pu

doi:10.1016/j.febslet.2007.09.018

more resistant than wild type to the formation of cellular or

cytotoxic edema after water intoxication [11].

Eclampsia is a common cause of maternal and foetal mortal-

ity during and after pregnancy, involving hypertension and

edema. A previous study reported immunoblot data indicating

strongly elevated AQP4 protein levels in pregnant and post-

partum rats, consistent with a pregnancy-induced upregulation

of AQP4 [12]. If confirmed, this would provide important in-

sight into the regulation of the AQP4 gene in particular, and

in water regulation in eclampsia in general.

Highly hydrophobic integral membrane proteins such as

AQP4 can behave atypically upon SDS–PAGE [13,14], the

electrophoresis technique of choice for quantitative measure-

ments of individual membrane proteins in tissue lysates. In this

study, we optimized procedures for obtaining reliable AQP4

SDS–PAGE immunoblot signals, with respect to detergent

choice, sample heating and gel composition, finding that these

factors all affect the quantity of AQP4 monomers.

By improved technique we have reinvestigated the important

claim of AQP4 upregulation in the brain of post-partum rats

[12]. Our results indicate no upregulation of neither AQP4 pro-

tein, nor AQP4 mRNA in the rat brain post-partum. We have

also investigated the reported upregulation of AQP4 in early

lung development [15], finding no significant upregulation of

AQP4. However, comparing AQP4 expression in the nose,

which has been reported in two previous studies [16,17], with

the reported upregulation of AQP4 in the cerebellum after

P7 [18], we find that AQP4 is upregulated as early in the nose

as in the brain. As this suggested a role for AQP4 in olfaction,

we investigated the expression pattern of AQP4 in the olfac-

tory bulb by immunohistochemistry, discovering that AQP4

is strongly expressed in the glomerulus, the multicellular syn-

aptic compartment of the olfactory bulb.

2. Materials and methods

2.1. Membrane fractionationMale PVG rats were killed by elevated CO2 and decapitated. The

brain was dissected and the cerebrum separated from the cerebellumand brain stem and put into chilled 50 mM HEPES–NaOH pH 7.4,2 mM EDTA, 0.32 M sucrose and protease inhibitor cocktail (Roche).The tissue was homogenized in a glass-homogenizer (Kontes), thencentrifuged 1000 · g 10 min, yielding the P1 (nuclear pellet) and S1(post-nuclear supernatant) fractions. S1 was subjected to anotherround of centrifugation at 164000 · g for 30 min using a BeckmanTi70 rotor. The resulting supernatant S2 (cytosolic fraction) was dec-anted and the pellet P2 (crude membrane fraction) was resuspendedin 10 ml of the same buffer and centrifuged once more under identicalconditions, yielding the P2 0 (washed membrane fraction). The P2 was

blished by Elsevier B.V. All rights reserved.

J.G. Sørbø et al. / FEBS Letters 581 (2007) 4884–4890 4885

resuspended in 5 ml of 50 mM HEPES pH 7.4, 2 mM EDTA and pro-tease inhibitor cocktail and stored at �20 �C.

2.2. Total fraction homogenatesRat brains from non-pregnant or post-partum (day 0–5) adult fe-

male Wistar rats were dissected and cerebellum/brain stem separatedfrom cerebrum. The tissue was homogenized in 1% SDS, 10 mM so-dium phosphate pH 7.4, 150 mM NaCl, 5 mM EDTA, protease inhib-itor cocktail and then sonicated. Homogenates were centrifuged1000 · g 10 min to pellet any insoluble material and yielding the totalfraction in the supernatant. One percent SDS solubilizes brain tissuecompletely as the homogenate goes into a clear solution after sonica-tion when used at appropriate tissue weight: buffer volume ratios.All fractions were assayed for total protein using the detergent-com-patible DC-Kit (Bio-Rad) with BSA diluted in H2O as a reference(the buffer alone gave only negligible readings).

2.3. AQP4 detergent solubilization assayFifty lg of mouse brain membranes were diluted to 1 g/L in 50 mM

HEPES pH 8.0, 1 mM DTT and one of a series of detergents: 60 mMCHAPS (Sigma), 5% v/v Triton X-100 (Electron microscopy sciences),5% DOC (Fluka), 5% v/v NP-40 substitute (Fluka), 5% v/v Decon90(Decon laboratories limited) or 1% w/v SDS (Fluka). Triton-X-100samples were sometimes supplemented with 150 mM NaCl or 8 Murea. In addition, 48 detergents from the Detergent Screening Kitfor Crystallization (Sigma–Aldrich) were tested (data not shown). Allsamples were incubated at room temperature with agitation for 1 h,then centrifuged at 23000 · g for 15 min. The resulting 50 ll superna-tant was mixed with 30 ll 6· loading buffer (SDS in excess was used toprevent a detergent competition effect), 45 ll loaded on gel. Pellet wasresuspended in 45 ll of 1· loading buffer, whole fraction was loaded.Gels were 10% SDS–PAGE for this assay.

2.4. SDS–PAGE and immunoblotting1· sample loading buffer was 1.7% SDS, 60 mM Tris pH 6.8, 5%

glycerol, and was supplemented with 100 mM DTT and 0.01% v/v b-mercaptoethanol where indicated. Molecular weight markers were See-bluePlus2 and MagicMarkXP (both Invitrogen). Two setups were usedfor electrophoresis and blotting: First, pre-cast 10% or 12% bis–trisgels and the mini-cell (Invitrogen) were used. Bis-tris gels were run withMOPS-SDS buffer (Invitrogen) at 200 V for �60 min or until the dyereached the bottom. Blotting was carried out according to the manu-facturer’s (Invitrogen) instructions using bis–tris blotting buffer with20% methanol and 0.2 lm PVDF (Bio-Rad). Immunodetection wasperformed as described below.

Second, the Mini-Protean III cell from Bio-Rad was used forself-cast gels according to Laemmli (16). Gels measured1.5 mm · 83 mm · 73 mm and the stacking gel was 15 mm. For con-ventional SDS–PAGE, gels consisted of 10% or 12% total acrylamidemonomer (C = 2.6%), 375 mM Tris pH 8.8, 0.1% SDS (resolving gel)and 4% or 5% total acrylamide monomer (C = 2.6%), 125 mM TrispH 6.8, 0.1% SDS. For Laemmli-Urea, gels were identical to the con-ventional gels with the exception of 3 M urea was included in both theresolving and the stacking gel. Urea concentrations from 2.8 M to 6 Mwere tested and no difference was found in terms of AQP4 monomerresolution. All self-cast gels were polymerized with the TEMED/per-sulfate system.

Self-cast gels were blotted with the Criterion cell (Bio-Rad). Briefly,after electrophoresis gels were equilibrated in Towbin buffer (25 mmTris, 192 mM glycine) supplemented with 20% methanol and blottedonto 0.2 lm PVDF (Bio-Rad) at 100 V for 30 min (increasing blottingtime to 90 min did not show improved transfer) with the same buffer.Successful transfer was approved by PonceauS staining (Sigma). Blotswere destained with H2O followed by 0.1 M NaOH, then blocked for15 min or more in 5% non-fat dried milk powder (Applichem) in TBST(20 mM Tris pH 7.6, 137 mM NaCl, 0.05% Tween20 (Sigma)) supple-mented with 0.05% NaN3.

Blots were incubated with rabbit-anti-AQP4 antibody (ChemiconAB3068) at a final concentration of 1 lg/ml in the blocking solutionovernight at 4 �C. Blots were washed in TBST, then incubated in alka-line phosphatase conjugated goat-anti-rabbit 1:10000 in TBST (Amer-sham, ECF Kit) 1 h at room temperature, washed for several hours inTBST and finally incubated 5 min with the ECF substrate before scan-ning on a Typhoon 9410 scanner (Amersham). Images were processed

using ImageQuantTL-software (Amersham). Blots were stripped andreprobed according to the manufacturer’s instructions (ECF Kit,Amersham). Antibodies for this purpose were mouse-anti-b-actin(Abcam ab8226) diluted 1:2000 and mouse-anti-GFAP (ChemiconmAB360) diluted 1:20000.

2.5. RNA isolation, cDNA synthesis and TaqMan real-time PCRFifty milligram pieces of tissues were dissected from animals killed

by carbon dioxide gassing and decapitation, and stored briefly inRNAlater (Ambion), before being homogenized and RNA extractedusing RNeasy (Qiagen). The RNA concentration was quantified usingNanoDrop (NanoDrop Technologies, Wilmington, USA) UV spec-trometry, and 2 lg RNA was converted into cDNA using High Capac-ity cDNA Archive Kit (Applied Biosystems). AQP4 expression wasmeasured using TaqMan assay Rn00563196_m1 (Applied Biosystems)on a 7900 HT Fast Real-Time PCR System (Applied Biosystems).

2.6. ImmunohistochemistryMale adult PVG rats were transcardially perfused with 25 ml 2%

dextran in 0.1 M Naphosphate (NaPi) buffer (pH 7.4), then perfusedwith 500 ml 4% paraformaldehyde in 0.1 M NaPi at 50 ml/min. Theteeth were excised and the nose region decalcified 24 h in 10% formicacid, then cryoprotected 24 h in 30% sucrose in 0.1 M NaPi. 16 lmcryosections were cut on a Leica CM3050 S cryostat. Sections wereincubated 20 min in blocking buffer (3% BSA, 0.5% TritonX-100,0.05% azide in TBS buffer), then incubated overnight at 4 �C with pri-mary antibody (goat-anti-AQP4, SantaCruz sc9888, 2 lg/ml) or non-immune goat IgG (Sigma I9140, 2 lg/ml) in blocking buffer. After3· 5 min washes in TBS, sections were incubated for 1 h in secondaryantibody (donkey-antigoat, Jackson #705-165-147, 1:1000 dilution) inblocking buffer. After 3· 5 min washes in TBS, the sections weremounted with Prolong Antifade Gold with DAPI (Invitrogen), andimages taken on a Zeiss LSM PASCAL Axioplan 2 Imaging confocalmicroscope. DAPI pictures were captures with the same microscope inepifluorescence mode.

3. Results and discussion

3.1. Detergent influence on AQP4 monomer signal

Agre and coworkers have proposed that AQP4 exists in vivo

as a heterotetramer [19], a hypothesis supported by recent

structural studies [20]. For reliable AQP4 monomer signal on

SDS–PAGE of AQP4, complete breakdown of tetramers is

essential. The choice of protein extraction parameters, in par-

ticular detergents, can play a key role in obtaining appropriate

AQP4 signals upon SDS–PAGE/immunoblotting. We devel-

oped a simple solubilization assay to measure the detergent ef-

fect on AQP4 protein extraction.

Membrane fractions, obtained by ultracentrifugation, were

incubated in various detergents, then centrifuged at

23000 · g, and the supernatant and pellet tested for presence

of AQP4. With no detergent present in the buffer, all AQP4

signal remained in the membrane pellet (P) after centrifuga-

tion, and no signal was solubilized to the supernatant (S)

(Fig. 1A, lanes 2 and 3).

Using CHAPS or Triton X-100, parts of the AQP4 signal is

observed in the supernatant (Fig. 1A, lanes 5 and 7). Interest-

ingly, while most of the AQP4 signal in the pellet is visible as

monomer, dimer and tetramer (Fig. 1A, lane 4 and 6, ‘‘1·’’,

‘‘2·’’, ‘‘4·’’), in the detergent solubilized fraction in the super-

natant, only the �120 kDa tetramer band is visible (Fig. 1A,

lanes 5 and 7, ‘‘4·’’).

The AQP4 tetramer can be seen even more clearly if the Tri-

ton solubilization mixture is supplemented with 150 mM NaCl

(Fig. 1A, lanes 8 and 9), or 8 M urea (lanes 10 and 11). Now

AQP4 is almost depleted in the pellet (lanes 8 and 10), and

Fig. 1. The effect of detergents on AQP4 solubilization and monomerresolution. (A) Lane 2–3: no detergent, 4–5: 60 mM CHAPS, 6–7: 5%Triton X-100, 8–9: 5% Triton X-100 + 150 mM NaCl, 10–11: 5%Triton X-100 + 8 M urea. (B) Lane 2–3: 5% DOC, 4–5: 5% NP-40substitute, 6–7: 5% Tween-20, 8–9: 5% Decon90, 10–11: 1% SDS.Putative monomer size 1·, dimer size 2· and tetramer size 4· indicatedon the right.

4886 J.G. Sørbø et al. / FEBS Letters 581 (2007) 4884–4890

AQP4 seems to be solubilized from the membranes into the

supernatant, but little or no AQP4 is seen as a monomer in

the supernatant (lanes 9 and 11). Similarly, DOC (deoxycho-

late) completely extracts AQP4 from the membrane to the

supernatant (Fig. 1B, lane 3), while NP-40-substitute only

gives a partial solubilization (Fig. 1B, lanes 4 and 5). Tween-

20 also provides a weak solubilization (Fig. 1B, lane 7).

Thus, if common lab detergents such as Triton X-100,

CHAPS or DOC were used in a lysate prepared for SDS–

PAGE, it could yield highly erroneous monomer signals or

no signal at all. The underlying reason why certain extraction

detergents make the AQP4 tetramer resistant to subsequent

breakdown to monomers in SDS–PAGE is unclear.

In contrast, the mixed non-ionic/anionic detergent Decon-90

gives an interesting partial solubilization from membranes,

and a supernatant with equally strong dimer and tetramer

bands in addition to the monomer (Fig. 1B, lane 9). Even with

SDS, which gives complete breakdown of the tetramer, one ex-

tra band at �60 kDa is seen in addition to the monomer

(Fig. 1B, lane 11, ‘‘1·’’), a band which is probably a dimer

(‘‘2·’’).

All other detergents tested (in total over 55 different deter-

gents were tested, data not shown) either did not solubilize

AQP4 completely or induced strong unspecific aggregation

that would hamper proper quantification of immunoblots.

Some detergents are better able to break down AQP4 tetra-

mers to dimers and monomers, but none of them, not even

SDS, the most widely used detergent, do manage a complete

breakdown of AQP4 monomers, as a strong dimer band re-

mains (Fig. 1B, lane 11). Thus, quantification of AQP4 expres-

sion is strongly dependent upon detergent choice.

The seemingly obscure in vitro detergent-dependent aggrega-

tion of AQP4 may have consequences for a rather large body

of literature on AQP4, and possibly may have consequences

for many other aquaporins and membrane proteins organized

as tetramers. Over 400 papers on AQP4 have been published

up to the present day, and many of the studies’ conclusions

might have been affected by the extraction procedure em-

ployed.

3.2. Temperature influence on AQP4 monomer signal

In early experiments with SDS-solubilized protein lysates

from post-partum pregnant rats, samples were vortexed and

heated at 90 �C before SDS–PAGE – in accordance with the

majority of published protocols. Most of the AQP4 signal col-

lapsed upon heating to a high molecular mass smear (Fig. 2A,

upper gel), while almost no signal was left at the expected

30 kDa monomer (Fig. 2A). The controls, the AQP4 isoforms

M1 and M23 were expressed in HeLa cells (Fig. 2A, lanes 8, 9,

16 and 17), only show weak signs of aggregation (Fig. 2A,

open triangle), indicating that brain lysate contains factors

aggravating aggregation.

To investigate the temperature dependency of AQP4 aggre-

gation quantitatively, a dilution series of samples (Fig. 2B,

lanes 4–9) were compared with heat treated samples to esti-

mate the signal loss. Samples prepared at 4 �C (lane 10),

20 �C (lane 11) and 37 �C (lane 12) show less monomer loss.

However, heating to 90 �C resulted in 90% loss of the AQP4

signal (Fig. 2, lane 13).

3.3. SDS–PAGE gel composition and AQP4 monomer

resolution

Previous investigations by Agre and coworkers indicated

that including urea in the gels increased the resolution of

AQP4 monomers upon SDS–PAGE/immunoblotting [19].

We tested the inclusion of urea concentrations from 2.8 M to

6 M in Laemmli gels. In particular, inclusion of urea in the

resolving gel improved AQP4 monomer resolution indepen-

dent of stacking gel composition, making possible the clear dis-

tinction of the M1 and M23 isoforms (Fig. 2C, enlarged in

Fig. 2D), while without urea, M1 and M23 isoforms cannot

be distinguished in the same lysates (Figs. 1A, B and 2A). Even

with urea included, smearing effects higher on the gels are per-

sistent (Fig. 2C). Furthermore, sample heating induces almost

complete loss of AQP4 monomer signal (Fig. 2C, lane 13).

Furthermore, we show that monomer resolution is unaffected

by the reducing agents beta-mercaptoethanol and DTT

(Fig. 2D).

3.4. Homodimerization of AQP4 M1 and M23

In several studies, attempts to resolve the �30 kDa AQP4

protein by SDS–PAGE apparently often resulted in additional

bands or smears appearing higher on the gels. Agre and

coworkers proposed already in 1996 that the �60 kDa bands

may be dimers of AQP4 [5]. This dimerization phenomenon

can be seen in many papers on mammalian Aqp4 [9,21–26],

Fig. 2. Temperature and urea dependency of AQP4 aggregation andresolution on SDS–PAGE/immunoblotting. (A) Severe aggregationdue to sample boiling. 12% bis–tris gel, 20 lg total protein per lane ofrat total brain samples; all samples heated to 90 �C prior to loading. (Band C) Improved isoform resolution with urea in the gel. All lanes rattotal brain lysate. All samples prepared at 20 �C, except lane 10 (onice), lane 12 (37 �C) and lane 13 (90 �C). Gel C was supplemented withurea. (D) Enlarged comparison of monomer separation without urea(lanes 1–3, left panel) and with urea (lanes 4–6, right panel). Samples inlanes 1 and 2, and lanes 4 and 5, were b-mercaptoethanol and DTT,respectively.

Fig. 3. (A) AQP4 homodimerization and dimerization of M23 inprimary culture. Lane 1 is size marker, with indicated sized in kDa.Lane 2 is M1 expressed in HeLa cells. Lane 3 is M23 expressed inHeLa cells. Lanes 4–10 are a dilution series of 0.16 lg, 0.32 lg, 0.5 lg,1 lg, 2 lg, 3 lg and 4 lg total hippocampus culture protein, respec-tively. (B) AQP4 in post-partum rats. Western blot with 20 lg totalprotein from cerebrum, from non-pregnant (NP) and post-partum(PP) rats. Lane 2: M1 and M23 controls expressed in HeLa cells, Lane3 and 4: NP, lane 5: PP0, lane 6: PP1, lane 7: PP3, lane 8: PP4 and lane9: PP5. (C) Comparison of early AQP4 expression in nose (N), cortex(C) and lung (L), at post-natal day P2, P4, P7 and P15. 30 lg proteinper lane.

J.G. Sørbø et al. / FEBS Letters 581 (2007) 4884–4890 4887

and has also been observed with AQP2 [27], and with DRIP,

the Drosophila aquaporin most similar to AQP4 [28].

AQP4 isoforms expressed in HeLa cells show a dimerization

tendency, with a corresponding shift in size of the dimers that

indicate homo-dimerization (Fig. 2A, lanes 8, 9, 16, 17, and

Supplementary Fig. 1). In primary mixed hippocampus/thala-

mus cultures, where neurons and glia have been cultured for

two weeks on glass slides, there is much less smear in the

SDS–PAGE gel signal, even if a dimer persists (Fig. 3A). Com-

paring dimer size with the M1 and M23 controls (Fig. 3A,

lanes 1 and 2) it is clear that the observed primary culture di-

mers of AQP4 Fig. 3A, lanes 4–10) are most similar to M23-di-

mers.

In a separate study we have biochemically fractionated

AQP4 from rat brain to purity. Even pure AQP4 exhibit

dimerization upon SDS–PAGE gels, and we have verified

by mass-spectrometry sequencing that the dimer is AQP4

(Sorbo et al., in preparation). How AQP4 is able to maintain

4888 J.G. Sørbø et al. / FEBS Letters 581 (2007) 4884–4890

a dimerized state in a denaturing gel environment is unclear.

We have shown that it is unlikely that dimerization is due to

covalent disulphide bridges (Fig. 2D).

3.5. AQP4 expression is not upregulated post-partum rat brains

Our optimized immunoblots were used to validate the claim

of a 9–22-fold upregulation of AQP4 in the post-partum rat

brain [12]. Using Laemmli/urea gels we obtained excellent

resolution of the two major isoforms M1 and M23, using

SDS-solubilized total fractions from cerebrum (Fig. 3B) and

cerebellum (data not shown) from five post-partum rats. The

densitometric analysis did not reveal any upregulation of brain

AQP4 protein in post-partum rats when compared to female

rat controls (data not shown). Even when post-partum samples

were compared to male rats, no difference in expression was

Fig. 4. Comparison of AQP4 expression in nasal epithelium and olfactory bwhole section including respiratory epithelium, olfactory epithelium and olfaletters. (B) Direct comparison of respiratory epithelium (left) and olfactorylamina propria and crebriform plate bone in upper left lumen (A). (D) Olfactlayer (right) (A). (E) Glomerulus in the glomerular layer of the medial left ollow background staining in glomerular layer of olfactory bulb.

found (data not shown). In order to supplement our protein

observations at the mRNA level, TaqMan RealTime RT-

PCR was employed. RNA was harvested from two post-par-

tum rats (day 1) and from two non-pregnant female control

rats. No significant change in the AQP4 expression level was

observed (data not shown).

3.6. Early AQP4 upregulation in rat nose mucosa, but not in

early lung development

AQP4 has previously been observed by freeze-fracture im-

muno-labeling (FRIL) of square arrays in the nasal epithelium

[16], by Western blot at post-natal day 21 in the distal lung and

nasopharynx of the adult rat [15], and recently, by immunohis-

tochemistry in nasal epithelium of the adult rat [17]. AQP4 has

been shown previously to be strongly upregulated in the brain

ulb by immunohistochemistry. (A) Dissection microscopy imaging ofctory bulb. Location of confocal images are indicated by red boxes and

epithelium (right) near lower septum (A). (C) Olfactory epithelium,ory nerve layer (left), glomerular layer (middle) and external plexiformfactory bulb (A). (F) Purified IgG primary antibody control show very

J.G. Sørbø et al. / FEBS Letters 581 (2007) 4884–4890 4889

between post-natal day 7 (P7) and P14 [18], and in the lung at

P2 [15]. Thus we wanted to test the sensitivity of our optimized

immunoblot by investigating the earliest upregulation of

AQP4 in the nose of rat pups, and compare it with the known

upregulation in the brain and the lung.

Interestingly, while finding no evidence of early AQP4

upregulation in rat lung (Fig. 3C, also negative in additional

data, n = 6 for P2, n = 3 for P1, data not shown), we found

that AQP4 was as early upregulated in the nasal mucosa as

in the brain (Fig. 3C). Since the strong upregulation of

AQP4 between P7 and P14 has been hypothesized to be vital

in the closing of the extracellular space and potassium homeo-

stasis in the brain [18], the parallel strong upregulation of

AQP4 in nose epithelium poses new questions about AQP4

function. Furthermore, the isoform profile in the rat nose

showed clearly the three distinct bands indicative of brain

AQP4, while the lung samples only showed a single, indistinct

band (Fig. 3C). We concluded that the early activation of

AQP4 in the nose pointed to a role for AQP4 in early olfactory

function in the rat pup.

3.7. Comparing AQP4 expression in the nasal epithelium and

glomerulus of the olfactory bulb

By immunohistochemistry, we could compare the signal

intensity in brain and nasal mucosa in a single section by using

coronal sections at the interface between the rostalmost part of

the olfactory bulb and the hindmost part of the nasal turbin-

ates and septum (Fig. 4A).

In confocal images from near the transition between respira-

tory and olfactory epithelium (Fig. 4A (B), we can obtain a di-

rect comparison of the strong AQP4 signal in the narrow layer

of respiratory epithelium (Fig. 4B, left), with the weak or ab-

sent AQP4 signal in the broad olfactory epithelium. In the

olfactory epithelium, however, the Bowman ducts, basal cells

and cells in the lamina propria are strongly labeled (Fig. 4B,

right). Our observations are consistent with a recent report

[17].

Similarly, in the respiratory lumen close to the crebriform

plate (Fig. 4A (C)), it can be seen that the AQP4 signal is pres-

ent in the lamina propria and Bowman’s ducts (Fig. 4C), but

there is little signal in the olfactory nerve fibers and the bone

(Fig. 4C, right), and in the olfactory epithelial cells (Fig. 4C,

left).

Very interestingly, while the olfactory nerve layer has little

or no AQP4 signal (Fig. 4D, left), a strong signal is present

in the glomerular layer of the olfactory bulb (Fig. 4D, middle).

In the external plexiform layer, only perivascular labeling

stands out (Fig. 4D, right). The strong AQP4 signal seemed

to correspond closely to the round glomerulus structures

(Fig. 4E). Since the glomerulus is the characteristic multicellu-

lar synaptic unit of the olfactory system, this again points to a

role for AQP4 in olfaction. Furthermore, since AQP4 has been

proposed to be involved in brain water homeostasis [8], in syn-

aptic function [29] and in cell migration [30], the presence of

AQP4 in the olfactory bulb glomerulus provides a novel sys-

tem for testing of alternative hypotheses of AQP4 function.

Acknowledgements: J.G.S. and S.E.M. have grants from Forskerlinjen,the Medical Student Research Programme at the University of Oslo.T.H. has a grant from The Norwegian Cancer Society. This workwas supported by the Nordic Centre of Excellence Program in Molec-ular Medicine and the Research Council of Norway (Storforsk Pro-gram).

Appendix A. Supplementary data

Supplementary data associated with this article can

be found, in the online version, at doi:10.1016/j.febslet.

2007.09.018.

References

[1] King, L.S., Kozono, D. and Agre, P. (2004) From structure todisease: the evolving tale of aquaporin biology. Nat. Rev. Mol.Cell Biol. 5, 687–698.

[2] van Balkom, B.W., van Raak, M., Breton, S., Pastor-Soler, N.,Bouley, R., van der, S.P., Brown, D. and Deen, P.M. (2003)Hypertonicity is involved in redirecting the aquaporin-2 waterchannel into the basolateral, instead of the apical, plasmamembrane of renal epithelial cells. J. Biol. Chem. 278, 1101–1107.

[3] van Balkom, B.W., Savelkoul, P.J., Markovich, D., Hofman, E.,Nielsen, S., van der, S.P. and Deen, P.M. (2002) The role ofputative phosphorylation sites in the targeting and shuttling of theaquaporin-2 water channel. J. Biol. Chem. 277, 41473–41479.

[4] Hasegawa, H., Ma, T., Skach, W., Matthay, M.A. and Verkman,A.S. (1994) Molecular cloning of a mercurial-insensitive waterchannel expressed in selected water-transporting tissues. J. Biol.Chem. 269, 5497–5500.

[5] Lu, M., Lee, M.D., Smith, B.L., Jung, J.S., Agre, P., Verdijk,M.A., Merkx, G., Rijss, J.P. and Deen, P.M. (1996) The humanAQP4 gene: definition of the locus encoding two water channelpolypeptides in brain. Proc. Natl. Acad. Sci. USA 93, 10908–10912.

[6] Yang, B., Ma, T. and Verkman, A.S. (1995) cDNA cloning, geneorganization, and chromosomal localization of a human mercu-rial insensitive water channel. Evidence for distinct transcriptionalunits. J. Biol. Chem. 270, 22907–22913.

[7] Turtzo, L.C., Lee, M.D., Lu, M., Smith, B.L., Copeland, N.G.,Gilbert, D.J., Jenkins, N.A. and Agre, P. (1997) Cloning andchromosomal localization of mouse aquaporin 4: exclusion of acandidate mutant phenotype, ataxia. Genomics 41, 267–270.

[8] Nielsen, S., Nagelhus, E.A., Amiry-Moghaddam, M., Bourque,C., Agre, P. and Ottersen, O.P. (1997) Specialized membranedomains for water transport in glial cells: high-resolution immu-nogold cytochemistry of aquaporin-4 in rat brain. J. Neurosci. 17,171–180.

[9] Nagelhus, E.A., Veruki, M.L., Torp, R., Haug, F.M., Laake,J.H., Nielsen, S., Agre, P. and Ottersen, O.P. (1998) Aquaporin-4water channel protein in the rat retina and optic nerve: polarizedexpression in Muller cells and fibrous astrocytes. J. Neurosci. 18,2506–2519.

[10] Verkman, A.S., Binder, D.K., Bloch, O., Auguste, K. andPapadopoulos, M.C. (2006) Three distinct roles of aquaporin-4in brain function revealed by knockout mice. Biochim. Biophys.Acta 1758, 1085–1093.

[11] Manley, G.T., Fujimura, M., Ma, T., Noshita, N., Filiz, F.,Bollen, A.W., Chan, P. and Verkman, A.S. (2000) Aquaporin-4deletion in mice reduces brain edema after acute water intoxica-tion and ischemic stroke. Nat. Med. 6, 159–163.

[12] Quick, A.M. and Cipolla, M.J. (2005) Pregnancy-induced up-regulation of aquaporin-4 protein in brain and its role ineclampsia. FASEB J. 19, 170–175.

[13] Laemmli, U.K. (1970) Cleavage of structural proteins during theassembly of the head of bacteriophage T4. Nature 227, 680–685.

[14] Towbin, H., Staehelin, T. and Gordon, J. (1979) Electrophoretictransfer of proteins from polyacrylamide gels to nitrocellulosesheets: procedure and some applications. Proc. Natl. Acad. Sci.USA 76, 4350–4354.

[15] King, L.S., Nielsen, S. and Agre, P. (1997) Aquaporins in complextissues. I. Developmental patterns in respiratory and glandulartissues of rat. Am. J. Physiol. 273, C1541–C1548.

[16] Rash, J.E., Davidson, K.G., Kamasawa, N., Yasumura, T.,Kamasawa, M., Zhang, C., Michaels, R., Restrepo, D., Ottersen,O.P., Olson, C.O., et al. (2005) Ultrastructural localization ofconnexins (Cx36, Cx43, Cx45), glutamate receptors and aquapo-rin-4 in rodent olfactory mucosa, olfactory nerve and olfactorybulb. J. Neurocytol. 34, 307–341.

4890 J.G. Sørbø et al. / FEBS Letters 581 (2007) 4884–4890

[17] Ablimit, A., Matsuzaki, T., Tajika, Y., Aoki, T., Hagiwara, H. andTakata, K. (2006) Immunolocalization of water channel aquapo-rins in the nasal olfactory mucosa. Arch. Histol. Cytol. 69, 1–12.

[18] Wen, H., Nagelhus, E.A., miry-Moghaddam, M., Agre, P.,Ottersen, O.P. and Nielsen, S. (1999) Ontogeny of water transportin rat brain: postnatal expression of the aquaporin-4 waterchannel. Eur. J. Neurosci. 11, 935–945.

[19] Neely, J.D., Christensen, B.M., Nielsen, S. and Agre, P. (1999)Heterotetrameric composition of aquaporin-4 water channels.Biochemistry 38, 11156–11163.

[20] Hiroaki, Y., Tani, K., Kamegawa, A., Gyobu, N., Nishikawa, K.,Suzuki, H., Walz, T., Sasaki, S., Mitsuoka, K., Kimura, K., et al.(2006) Implications of the aquaporin-4 structure on arrayformation and cell adhesion. J. Mol. Biol. 355, 628–639.

[21] Terris, J., Ecelbarger, C.A., Marples, D., Knepper, M.A. andNielsen, S. (1995) Distribution of aquaporin-4 water channelexpression within rat kidney. Am. J. Physiol. 269, F775–F785.

[22] Vajda, Z., Promeneur, D., Doczi, T., Sulyok, E., Frokiaer, J.,Ottersen, O.P. and Nielsen, S. (2000) Increased aquaporin-4immunoreactivity in rat brain in response to systemic hyponatre-mia. Biochem. Biophys. Res. Commun. 270, 495–503.

[23] Silberstein, C., Bouley, R., Huang, Y., Fang, P., Pastor-Soler, N.,Brown, D. and Van Hoek, A.N. (2004) Membrane organizationand function of M1 and M23 isoforms of aquaporin-4 in epithelialcells. Am. J. Physiol. Renal Physiol. 287, F501–F511.

[24] Kim, S.W., Cho, S.H., Oh, B.S., Yeum, C.H., Choi, K.C., Ahn,K.Y. and Lee, J. (2001) Diminished renal expression of aquaporin

water channels in rats with experimental bilateral ureteralobstruction. J. Am. Soc. Nephrol. 12, 2019–2028.

[25] Huang, Y., Tracy, R., Walsberg, G.E., Makkinje, A., Fang, P.,Brown, D. and Van Hoek, A.N. (2001) Absence of aquaporin-4water channels from kidneys of the desert rodent Dipodomysmerriami merriami. Am. J. Physiol. Renal Physiol. 280, F794–F802.

[26] Yang, B., Van Hoek, A.N. and Verkman, A.S. (1997) Very highsingle channel water permeability of aquaporin-4 in baculovirus-infected insect cells and liposomes reconstituted with purifiedaquaporin-4. Biochemistry 36, 7625–7632.

[27] Shinbo, I., Fushimi, K., Kasahara, M., Yamauchi, K., Sasaki, S.and Marumo, F. (1999) Functional analysis of aquaporin-2mutants associated with nephrogenic diabetes insipidus by yeastexpression. Am. J. Physiol. 277, F734–F741.

[28] Kaufmann, N., Mathai, J.C., Hill, W.G., Dow, J.A., Zeidel, M.L.and Brodsky, J.L. (2005) Developmental expression and biophys-ical characterization of a Drosophila melanogaster aquaporin.Am. J. Physiol. Cell Physiol. 289, C397–C407.

[29] Amiry-Moghaddam, M. and Ottersen, O.P. (2003) The molecularbasis of water transport in the brain. Nat. Rev. Neurosci. 4, 991–1001.

[30] Saadoun, S., Papadopoulos, M.C., Hara-Chikuma, M. andVerkman, A.S. (2005) Impairment of angiogenesis and cellmigration by targeted aquaporin-1 gene disruption. Nature 434,786–792.