Báo cáo khoa học: Proteolysis of the tumour suppressor hDlg in response to osmotic stress is mediated by caspases and independent of phosphorylation pot

We first checked the regulation of hDlg phosphorylation by hyperosmotic stress at different cell densities using a phosphospecific antibody, which recognizes phosphorylated serine 158 S158

osmotic stress is mediated by caspases and independent

of phosphorylation

Francisco A In˜esta-Vaquera1, Francisco Centeno2, Paloma del Reino1, Guadalupe Sabio3, Mark Peggie3and Ana Cuenda1,3

1 Departamento de Inmunologı´a y Oncologı´a, Centro Nacional de Biotecnologı´a-CSIC, Madrid, Spain

2 Departamento Bioquı´mica y Biologı´a Molecular, Universidad de Extremadura, Ca´ceres, Spain

3 MRC Protein Phosphorylation Unit, University of Dundee, UK

Mammalian cells respond to changes in the osmolarity

of the medium by activating multiple signalling

path-ways, with p38 mitogen-activated protein kinases

(MAPKs) critical for both early response and long-term

cellular adaptation to prolonged hyperosmotic exposure

[1] Although all four p38 MAPKs (p38a, p38b, p38c

and p38d) are activated in response to hyperosmotic

stress, activation of the isoform p38c is particularly

rapid and strong compared with other p38s [2,3]

Recently, we described a novel regulatory pathway for

the adaptation of cells to a hyperosmolar environment

that acts parallel to the classical p38a pathway, which

involves the protein kinase p38c and its substrate hDlg,

and modulates the composition of the cytoskeletal

protein by phosphorylating one of its components [3]

hDlg (also called Dlg1 and dlgh1) is the human orthologue of the Drosophila tumour suppressor Dlg, and belongs to the membrane-associated guanylate kinase family of scaffold proteins, whose members have a similar structural organization composed of a basic core of a variable number of PDZ domains, a SH3 domain and a catalytically inactive guanylate kinase-like region [4] Functions of hDlg are related to the establishment and maintenance of cell polarity and the adhesion integrity of intestinal epithelial cells [5,6] Moreover, gene-targeted mice lacking full-length hDlg show defects in the morphogenesis of the kidney and urogenital tracts [7,8]

Evidence suggests that alterations in hDlg function may contribute to the development of cancer The

Keywords

apoptosis; caspase; human disc-large;

osmotic shock; p38-mitogen activated

protein kinase

Correspondence

A Cuenda, Departamento de Inmunologı´a y

Oncologı´a, Centro Nacional de

Biotecnologı´a-CSIC, Campus de

Cantoblanco-UAM, 28049-Madrid, Spain

Fax: +34 91 372 0493

Tel: +34 91 585 5451

E-mail: acuenda@cnb.uam.es

(Received 6 August 2008, revised 29

October 2008, accepted 7 November 2008)

doi:10.1111/j.1742-4658.2008.06783.x

Human disc-large (hDlg) is a scaffold protein critical for the maintenance

of cell polarity and adhesion hDlg is a component of the p38c MAP kinase pathway, which is important for the adaptation of mammalian cells

to changes in environmental osmolarity Here we report a strong decrease

in the levels of hDlg protein in the human epithelial cell line HeLa when exposed to osmotic shock This is independent of the phosphorylation state

of hDlg, is prevented by preincubating the cell with the caspase inhibitor z-VAD and is part of the apoptotic process triggered by cellular stress Although, both caspase 3 and caspase 6 are strongly activated by osmotic shock, the time course of caspase 6 activation parallels hDlg degradation, suggesting that this caspase may be responsible for the proteolysis Mutat-ing hDlg Asp747 to Ala abolishes caspase-induced cleavage, but does not affect the early stage of apoptosis or cell attachment Our findings show that osmotic stress triggers hDlg degradation through a mechanism differ-ent from the one mediated by proteasomes, and we iddiffer-entify hDlg as a caspase substrate during the apoptotic process, although its proteolysis may not be implicated in the progression of early apoptosis

Abbreviations

GST, glutathione S-transferase; hDlg, human disc large; HPV, human papillomavirus; MAPK, mitogen-activated protein kinase; PSI,

proteasome inhibitor I.

expression of hDlg in epithelial-derived cancers (such

as cervical, gastric and colon cancers) is extremely low

or even absent [9], in addition, hDlg binds to

oncopro-teins expressed by viruses such as, human

papillomavi-rus (HPV), human T-cell leukaemia vipapillomavi-rus type 1 and

human adenovirus type 9 [10] The binding of HPV E6

protein to hDlg causes a decrease in hDlg protein

levels by inducing its proteasome-mediated degradation

[11] In epithelial cell lines, this degradation is highly

dependent on the state of hDlg phosphorylation and

the degree of isolation of the cell

Hyperphosphoryla-tion of hDlg makes it more susceptible to degradaHyperphosphoryla-tion

induced by the HPV E6 oncoprotein [12], whereas in

isolated cells, infected or not with HPV, the

degrada-tion of hDlg is constitutive in the cytoplasm [13]

In addition to the suggestion that hDlg

phosphory-lation may modulate its protein levels in cells, in recent

years, phosphorylation has emerged as a mechanism

for regulating hDlg’s function as a scaffold protein

[5,14,15] Accordingly, we have shown that hDlg is

hyperphosphorylated in response to cellular stress such

as osmotic shock or UV radiation This

phosphoryla-tion is mediated by p38c MAPK and triggers its

dissociation from the cytoskeletal protein GKAP,

therefore releasing it from the cytoskeleton into the

cytoplasm [3]

Our aim in this study was to gain a better

under-standing of the role of hDlg phosphorylation by

p38c when cells are exposed to hyperosmotic stress

Here we analyse whether the phosphorylation of hDlg,

triggered by osmotic shock, could also control levels of

hDlg protein in the human epithelial cell line HeLa We

report a strong decrease in hDlg protein, although this

event is independent of its phosphorylation state

More-over, this hDlg proteolysis was dependent on caspase

activation during the apoptosis process in the cells

Results

Osmotic shock causes a decrease of hDlg protein

As mentioned previously, hDlg degradation seems to

be regulated by phosphorylation and cell density

[12,13,16] Moreover, we have reported that when cells

are exposed to osmotic shock, endogenous hDlg is

hyperphosphorylated by the protein kinase p38c [3]

Therefore, we initiated experiments to determine

whether hDlg degradation is affected by

phosphory-lation mediated by p38c in a cell density-dependent

manner We first checked the regulation of hDlg

phosphorylation by hyperosmotic stress at different

cell densities using a phosphospecific antibody, which

recognizes phosphorylated serine 158 (S158), a hDlg

residue that becomes phosphorylated by p38c follow-ing sorbitol treatment [3,17] As expected, we observed

an increase in hDlg phosphorylation upon sorbitol stimulation under all experimental conditions (Fig 1A) Moreover, a significant increase in basal hDlg phosphorylation ( 26%) was observed in con-fluent cells, but this may be due to an increase in the total amount of hDlg protein ( 30%) when cells were 100% confluent, consistent with the stabilization of this protein upon cell–cell contact (Fig 1A) [16] Once we had confirmed that hDlg was phosphory-lated, we examined the effect of osmotic stress on hDlg protein levels and whether this was dependent on cell density To avoid cell death caused by long exposure

to hyperosmotic shock, cells were treated for 60 min with sorbitol and then released into fresh medium for

9 or 14 h Under these conditions, we detected a large decrease in hDlg protein levels, of 75% at 9 h and 85% at 14 h after the treatment with sorbitol ceased Moreover, hDlg loss was similar in cells 50 or 100% confluent, indicating that cell density does not affect the decrease in hDlg protein caused by osmotic shock (Fig 1B) We found that the levels of hDlg in HeLa cells treated with sorbitol decrease progressively with time after stress, falling to 30% at 6 h after release from sorbitol treatment (Fig 1C) For comparison, exposure of cells to UV radiation, which also triggers hDlg phosphorylation mediated by p38c as well as its degradation [3,18], causes a decrease in hDlg protein levels similar to that observed after osmotic sock treat-ment (Fig 1D) No strong stable accumulation of hDlg cleavage fragments could be detected at these time points (not shown), indicating that either the deg-radation of this molecule is very rapid and at multiple sites or the antibodies used do not recognize the epitope of the cleavage fragments

Osmotic shock-induced degradation of hDlg is mediated by caspase

We measured the activation of different proteases after hyperosmotic shock or UV treatment in HeLa cells to determine which might be involved in hDlg degrada-tion As shown in Fig 2A, only caspase 3 and cas-pase 6 activities were significantly induced Other protease activities, including caspase 8, calpain or cathepsin B, were not affected Activation of the effec-tor caspase 3 and caspase 6 suggests that hDlg could

be proteolysed by one of them To check this, HeLa cells were treated with or without the general caspase inhibitor z-VAD prior to and during exposure to stress As shown in Fig 2B, z-VAD blocked hDlg degradation and the activation of caspase 3 and

caspase 6 (Fig 2C) Other protease inhibitors, such as

the calpain inhibitor MDL 28170 or the cathepsin

inhibitor 3-Met adenine, did not affect hDlg

degrada-tion (data not shown) Because it has been shown that

the reduction in hDlg levels may be proteasome

depen-dent, we also treated the cells with the proteasome

inhibitor proteasome inhibitor I (PSI), and found that

this did not prevent hDlg degradation (Fig 2B),

although it did prevent degradation of the kinase

SGK1, the steady-state level of which is very low

because of its degradation by proteasomes (Fig 2D)

[19] These results indicate that the degradation of

hDlg upon cellular stress is mediated by caspases and

not by the proteasome

In addition, when we examined the time course of

activation of caspases by cellular stress, we found that

activation of caspase 6 is greater and more sustained

than that of caspase 3 (Fig 2E) Moreover, caspase 6

activation reached its maximum at 3 or 6 h after UV

or sorbitol treatment, respectively, and then decreased

Whereas caspase 3 was transiently activated, being

maximal 30–60 min after exposure to either stimulus These results show that the loss of hDlg over the time course of sorbitol or UV treatment (Fig 1C,D) corre-lated with the kinetics of caspase activation, particu-larly of caspase 6, supporting that hDlg cleavage is caspase dependent

Phosphorylation by p38c does not modulate hDlg proteolysis

To determine whether hDlg degradation induced by osmotic stress was dependent on phosphorylation by p38c, we first expressed wild-type glutathione S-trans-ferase (GST)–hDlg, which behaves similarly to endoge-nous hDlg (Fig S1), or different GST–hDlg mutants,

in which each of the in vivo p38c-phosphorylation sites were mutated individually to Ala to prevent phosphorylation (S158A, T209A, S431A and S442A) [3] We found that the amount of GST–hDlg wild-type and of the different GST–hDlg mutants decreased equally following osmotic shock treatment (Fig 3A)

A

hDlg (S158)

hDlg (total)

B

hDlg (total)

GAPDH

Time (h)

0 9 14

– –

+

0 9 14

hDlg GAPDH

0 20 40 60 80 100

GAPDH hDlg

Time after sorbitol treatment (h)

Time after UV treatment (h)

9

D

0 20 40 60 80 100

C

Fig 1 Osmotic shock causes a decrease in

hDlg protein levels in a manner not

depen-dent on cell density (A) HeLa cells were

exposed for 15 min to 0.5 M sorbitol.

Endogenous hDlg was immunoprecipitated

from 0.4 mg of cell lysate and pellets were

immunoblotted using an antibody that

recognizes phosphorylated hDlg [hDlg

(S158)] or total hDlg (anti-hDlg) (B) Cells

were exposed to hyperosmotic stress

(0.5 M sorbitol) for 60 min, and then

released into sorbitol-free medium for 0, 9

or 14 h Levels of hDlg were analysed by

immunoblot using hDlg Ig (C) (Upper) HeLa

cells treated as in (B) Endogenous levels of

hDlg were analysed, by immunoblot with

hDlg Ig, 1, 3 and 6 h after sorbitol had

been washed out (Lower) hDlg levels were

quantified as described in Materials and

methods Values are means (± SE) of three

independent experiments (D) HeLa cells

were exposed to UV irradiation (200 JÆm)2),

followed by 3, 6, 9 or 14 h incubation.

Endogenous levels of hDlg were analysed

as in (C) Quantification values are means

(± SE) of two independent experiments.

Immunoblots are shown as one

representa-tive experiment Endogenous GAPDH level

was used as a loading control Lines in this

figure are duplicates.

These results suggest that the phosphorylation of

indi-vidual sites does not modulate hDlg degradation and

prompted us to examine whether it was controlled by

phosphorylation at more than one site We then per-formed the same experiment using a hDlg mutant [GST–hDlg(A6)] in which all sites phosphorylated by

Caspase 3 1.25 + 0.068 13.22 + 0.35 6.11 0.53

Caspase 6 1.71 + 0.26 7.74 + 1.11 4.73 +

+

+ +

1.17 Caspase 8 1.43 + 0.03 1.67 + 0.13 1.82 0.23

Calpain 0.33 + 0.03 0.55 + 0.01 0.8 0.05

CathepsinB 0.61 + 0.08 0.47 + 0.05 0.36 + 0.02

0 3 6 9 12

0 5 10 15 20 25

Time after treatment (h)

Sorbitol

UV-C

Caspase 6

Sorbitol

UV-C

Caspase 3 E

Sorbitol

zVAD PSI

–

+

+ –

+

+

– – –

–

–

– – –

hDlg GAPDH

– –

–

– –

–

– –

–

–

+

+ +

B

0 25 50 75 100

0 5 10 15

Sorbitol

zVAD UV-C

+ + +

-–

Caspase 3

+ + +

-Caspase 6

50

37

SGK1

D A

C

p38c in vitro and in vivo were mutated to Ala and

therefore could not be phosphorylated under stress

The mutated phosphorylation sites were the four sites

regulated by stresses (S158, T209, S431 and S442), one

site constitutively phosphorylated in cells (S122) and

one site phosphorylated in vitro by p38c (S447) [3]

Surprisingly, both hDlg wild-type and mutant

disap-peared to a similar extent in cells treated with sorbitol

(Fig 3B), indicating that phosphorylation of hDlg by

p38c does not regulate its degradation induced by

osmotic stress in HeLa cells

To confirm these findings and verify whether hDlg

degradation was dependent on its state of

phosphoryla-tion, we treated cells with different MAPK pathway

inhibitors to block hDlg phosphorylation [17] We then

examined hDlg loss in cells treated with inhibitors and

exposed to sorbitol Both p38MAPK inhibitors,

BIRB0796, which at high concentrations inhibits all p38

and JNK isoforms [17], and SB203580, which inhibits

the isoforms p38a⁄ b, failed to abolish hDlg degradation (Fig 3C) BIRB0796 at high concentrations (1 and

10 lm), but not SB203580, efficiently blocked hDlg phosphorylation [17] (data not shown) Because osmotic stress also may cause the activation of other MAPKs such as ERK1⁄ 2, ERK5 or JNK, we investigated whether these kinases were involved in hDlg degrada-tion Treatment of cells with PD184352 at low concen-tration abolishes the activation of ERK1⁄ 2 and at high concentrations abolishes the activation of ERK5; treat-ment with SP600125 blocks JNK activity along with other many protein kinases such as SGK1, PRAK, AMPK, CHK, CDK2 or S6K1 [20] None of these inhibitors blocked the decrease in hDlg protein levels (Fig 3C) triggered by osmotic shock (or UV, data not shown), although they inhibited the different MAPKs activations (Fig S2) These results suggest that other MAPK family members activated by cellular stress do not control the disappearance of hDlg

Fig 2 hDlg degradation is mediated by caspases (A) HeLa cells were treated with 0.5 M sorbitol for 60 min or with UV irradiation (200 JÆm)2), followed by 1 h incubation in stimulus-free media, proteases activity was determined by fluorescence emitted from specific peptide cleavage, as described in Materials and methods Values are means (± SE) of three independent experiments (B) Cells were prein-cubated in the absence or in the presence of 30 l M caspase inhibitor z-VAD or 60 l M proteasome inhibitor PSI, 60 min before treatment with sorbitol or UV irradiation, as in (A), followed by 14 h incubation in stimulus-free media (in the continued absence or presence of prote-ase inhibitor) Endogenous levels of hDlg were analysed and quantified as in Fig 1C (C) HeLa cells were preincubated with (grey bars) or without (black bars) the pan-caspase inhibitor z-VAD for 1 h before treatment with sorbitol or UV irradiation, as in (A) Values are means (± SE) of three independent experiments (D) HeLa cells were incubated with or without 60 l M PSI and endogenous levels of SGK1 were analysed by western blotting using SGK1 Ig (E) HeLa cells (circles) or HEK293 cells (triangles) were exposed, as in (A), to sorbitol (black circles or triangles) or to UV irradiation (empty circles or triangles), then released from the stimuli for the times indicated before assaying caspase activities Values are means (± SE) of two independent experiments.

C

Sorbitol BIRB0796 (µ M ) SB203580 (µ M ) PD184352 (µ M ) SP600125 (µ M )

– – –

– – – 10

– –

–

– – –

– – – –

– – –

– – –

– – –

– – –

– – – – –

10 – – 2 –

– –

– 10 –

– – – – –

hDlg GAPDH

Fig 3 The decrease in hDlg protein level is not regulated by p38c phosphorylation (A) HeLa cells were transfected with either GST–hDlg wild-type or different GST–hDlg mutants in which in vivo p38c phosphorylation sites had been mutated to Ala, and then exposed 0.5 M sorbi-tol for 60 min Overexpressed GST–hDlg were analysed 6 h after sorbisorbi-tol had been washed out, as in Fig 1 (B) Cells were transfected with GST–hDlg wild-type or GST–hDlg(A6) mutant Cells were treated as in (A) and GST–hDlg was analysed 1, 3 and 6 h after sorbitol release Lines indicate duplicates (C) Cells were preincubated for 2 h with or without different kinase inhibitors, at the concentrations indicated Cells were exposed to 0.5 M sorbitol for 60 min, and then incubated for 9 h in sorbitol-free medium (in the continued absence or presence of inhibitor) Endogenous levels of hDlg were analysed as in Fig 1C GAPDH or p38a levels were used as loading controls.

To determine whether hDlg phosphorylation

regu-lates hDlg degradation and to exclude the possibility

that other phosphorylation sites, different from those

for p38c, may modulate hDlg degradation, we treated

cells with the general kinase inhibitor staurosporine

[21] To test whether staurosporine could inhibit hDlg

phosphorylation, cells were preincubated with or

with-out this compound for 1 h before exposure to sorbitol

Phosphorylation of endogenous hDlg was blocked

completely by 1 lm staurosporine (Fig 4A) However,

preincubation of cells with staurosporine failed to

block the decrease in hDlg protein level, and caused a

significant increase in hDlg degradation (Fig 4B)

Given the above results, we decided to check whether

incubation of cells with staurosporine alone caused

hDlg loss As shown in Fig 4C, staurosporine also

caused hDlg degradation in a concentration- and

time-dependent manner

hDlg is degraded in apoptotic cells

These findings suggest that hDlg degradation might be

related to the apoptosis of the cell, because

stauro-sporine is a potent inducer of caspase-dependent cell

apoptosis [21] (data not shown) Therefore, we

estab-lished that, after cellular stress, cells undergo

apopto-sis HeLa, or HEK293 cells for comparison, were

exposed to either sorbitol or UV, and apoptosis was

determined 3 or 14 h after exposure to the stimulus

As shown in Fig 5A, 70% of HeLa cells underwent

apoptosis 14 h post stimulus, whereas only  15% of

HEK293 cells started to die 14 h after sorbitol or UV

treatment Basal levels of apoptosis in control cells,

not exposed to stress, were also significantly higher in HeLa cells than in HEK293 cells (Fig 5A) Accord-ingly, when we examined the levels of hDlg protein in HEK293 cells that had been exposed to sorbitol or

UV, we could not detect any protein degradation even

14 h after the stimulus (Fig 5B) However, neither

0 25 50 75 100

Staurosporine (µM)

2

GAPDH hDlg

0 Staurosporine (µM)

Time (h)

C

B

0 25 50 75 100

Staurosporine (µM)

Sorbitol

GAPDH hDlg

Staurosporine (µM)

Sorbitol

hDlg (S158)

hDlg (total)

A

hDlg (T209) hDlg (S431) hDlg (S442)

Fig 4 The broad-spectrum kinase inhibitor staurosporine enhances

hDlg degradation (A) Cells were incubated for 1 h with or without

the indicated concentration of staurosporine and then exposed for

15 min to 0.5 M sorbitol hDlg was immunoprecipitated and

analy-sed using antibodies that recognize phosphorylated hDlg at four

different residues, hDlg (S158), hDlg (T209), hDlg (S431) or hDlg

(S442), or an antibody that recognizes total hDlg (anti-hDlg) (B)

HeLa cells were preincubated for 1 h with or without the

pan-kinase inhibitor staurosporine at the concentrations indicated, and

treated with 0.5 M sorbitol for 60 min, and then incubated for 3 h in

sorbitol-free medium (in the continued absence or presence of the

inhibitor) Endogenous levels of hDlg were analysed by

immuno-blotting (lower) and the percentage of protein level quantified

(upper) as described in Materials and methods Values are means

(± SE) of three independent experiments (C) HeLa cells were

trea-ted without or with 1 or 10 l M staurosporine for 2 h (upper and

lower) or 7 h (lower) Endogenous levels of hDlg were analysed

as described previously and the percentage of hDlg protein was

quantified (lower) Values are means (± SE) of two independent

experiments.

caspase 3 nor caspase 6 were activated in HEK293

cells under any of the experimental conditions

described (Fig 2E), which is consistent with the lack

of hDlg degradation observed in this cell line The low

percentage of HEK293 cells undergoing apoptosis

(compared with HeLa cells), and the the lack of hDlg

loss in HEK293 cells, support the idea that hDlg

deg-radation in HeLa cells is part of the apoptotic event

triggered by cellular stress

Changes in hDlg localization upon hyperosmotic

shock exposure

Although hDlg is normally localized in adherens

junc-tions at sites of cell–cell contact, its cellular

distribu-tion is different in confluent and subconfluent cells,

hence our next question was whether this localization

would change during apoptosis induced by sorbitol As

expected, when HeLa cells were 50% confluent, hDlg

was localized diffusely throughout the cytosol and at

the membrane hDlg localization changed when cells

reached confluency; in this condition, hDlg was present

mainly at the membrane, whereas the amount found in

the cytoplasm decreased markedly (Fig 6A) These

results were confirmed by subcellular fractionation

analysis; we found that the amount of hDlg in the

cytoplasm is greater in subconfluent cells, but in

confluent cells hDlg is mainly in the membrane frac-tion (Fig 6B) After exposing the cells to sorbitol (Fig 6A,B), the total amount of hDlg decreased in both 50% and 100% confluent cells, and this decrease was equal in all cell compartments (Fig 6B) These results show that hDlg localizes to the plasma mem-brane when cells reach confluency and establish cell– cell contact, and that its degradation occurs in all cell compartments in which hDlg is present

In addition, in CaCo-2 cells, derived from human colonic adenocarcinoma and in which hDlg is degraded in response to osmotic stress (Fig S3), hDlg was found mostly in areas of cell–cell contact and a substantial and gradual loss from the mem-brane was observed after osmotic shock treatment (Fig 6C) The more compact localization of hDlg is partially lost and hDlg is localized more diffusely throughout the cytoplasm in the vicinity of the mem-brane (Fig 6Ca–c) However, pretreatment of the cell with the p38MAPK inhibitor BIRB0796 causes a sig-nificant delay in the loss of Dlg from the membrane (Fig 6Cd–f), whereas pretreatment with the caspase inhibitor z-VAD, in both the absence and presence

of BIRB0796, preserved the membrane localization

of hDlg under all experimental conditions (Fig 6Cg– l) Because cell–cell contacts are largely preserved at these initial time points, it is suggested that hDlg

Sorbitol

0 20 40 60 80

UV-C

Time after stimuli treatment (h)

A

Sorbitol

UV-C

hDlg GAPDH hDlg GAPDH

Fig 5 Cellular stresses induce apoptosis in

HeLa cells (A) HeLa (black bars) or HEK293

cells (grey bars) were exposed to 0.5 M

sorbitol for 60 min or to UV irradiation

(200 JÆm)2) followed by 3 or 14 h incubation

in stimulus-free media before quantitative

analysis of apoptosis by propidium iodide ⁄

annexin V Values are means (± SE) of three

independent experiments (B) HEK293 cells

were exposed to 0.5 M sorbitol for 60 min

or to UV irradiation (200 JÆm)2), followed by

0, 3, 6 or 14 h incubation Endogenous

levels of hDlg were analysed as described

in Fig 1.

Control

Sorbitol

Cell density

Control

Sorbitol

hDlg hDlg + DAPI hDlg hDlg + DAPI

hDlg hDlg + DAPI hDlg hDlg + DAPI

hDlg InsR Calpain

Cell density 50% 100%

B

Treatment

None

1 h Sorbitol

1 h Sorbitol

+

3 h release

BIRB0796 + zVAD

C

a

b

c

d

e

f

g

h

i

j

k

l BIRB0796

cleavage represents a step in apoptosis that may

precede cell–cell detachment

Identification of hDlg caspase-cleavage sites

These findings suggest that hDlg is a caspase target

To confirm this and identify the caspase(s) cleavage

site(s) on hDlg, experiments were performed using

mutations that affect putative caspase 3 and⁄ or

cas-pase 6 sites Within their substrates, cascas-pases

recog-nize a core tetrapeptide motif (P4P3P2P1) that

contains an essential aspartic acid residue required

for the cleavage reaction at position P1 [22] One

well-defined substrate-recognition motif for caspases,

in particular caspase 3, is DXXD, whereas for

cas-pase 6 it is (I⁄ V ⁄ L)EXD [22] Analysis of the hDlg

protein sequence indicates that it contains several

putative caspase-cleavage sites, DVRD255 and

DGRD750, which are typical caspase 3 recognition

sequences, YEVD747, which may be another potential

caspase site, and QSVD397, which has been

previ-ously reported as an unusual caspase 3 cleavage site

[18] We generated different hDlg mutants in which

the aspartic acid residue predicted as being required

for caspase cleavage was mutated to Ala (Fig 7A)

These constructs were transfected to HeLa cells and

the effects of sorbitol (and UV) treatment on GST–

hDlg wild-type or mutant GST–hDlg degradation

were compared As shown in Fig 7B, 3 h after

stim-ulation ceased, only the mutation D747A blocked

the degradation of GST–hDlg, although wild-type

hDlg and the other hDlg mutant forms were

degraded to the same extent after treatment

Quanti-fication of hDlg protein confirmed that D747A was

the only mutant not cleaved after cellular stress

(Fig 7C) These results identify the sequence YEVD

as a possible site of hDlg cleavage in early apoptotic

cells

Effect of hDlg cleavage on cell–cell detachment

and early stage of apoptosis

To evaluate the role of hDlg during these processes,

HeLa cells were transfected with hDlg wild-type or

mutant hDlgD747A, which is not degraded, or mutant

hDlgD397A, as a control We tried to generate cell lines stably overexpressing wild-type or hDlg mutants However, none of the attempts was successful There-fore, the transfection procedure was optimized to

Control Sorb UV-C

Tubulin

Tubulin

Tubulin Tubulin Tubulin

hDlg(WT)

hDlg(D255A)

hDlg(D397A)

hDlg(D255A/D750A)

hDlg(D747A)

B

C

0 20 40 60 80 100

A

PDZ1 PDZ2 PDZ3 SH3 GUK

hDlg

Fig 7 Identification of the hDlg-caspase-cleavage site (A) Sche-matic representation of hDlg and the putative caspase-cleavage sites (B) Cells were transfected with GST–hDlg wild-type or differ-ent GST–hDlg mutants, in which differdiffer-ent caspase-cleavage sites have been mutated to Ala, and then exposed to 0.5 M sorbitol for

60 min or UV irradiation (200 JÆm)2) GST–hDlg were analysed 3 h after UV treatment or after sorbitol had been washed out, as in Fig 1 The endogenous tubulin level was use as the loading con-trol Immunoblots are shown as one representative experiment (C) Percentage GST–hDlg protein level was quantified as before: hDlg protein from control cells (black bars), cells treated with sorbitol (dark grey bars) or treated with UV (light grey bars) Values are means (± SE) of three to four independent experiments.

Fig 6 Localization of hDlg in apoptotic cells (A) HeLa cells were grown at 50 or 100% confluency, exposed or not to 0.5 M sorbitol for

60 min, stained with hDlg Ig 3 h after sorbitol release, and subjected to immunofluorescence microscopy Nuclei are stained with DAPI Similar results were obtained in three independent experiments (B) HeLa cells were exposed to 0.5 M sorbitol for 60 min, and then released

in sorbitol-free medium for 0 or 9 h Cells were subjected to cellular fractionation as indicated in Materials and methods and 10–30 lg of protein from cytoplasm and membrane fractions were immunoblotted using the antibodies indicated (C) CaCo-2 cells were preincubated with or without the kinase inhibitor BIRB0796 (1 l M ) and in the presence or absence of z-VAD (30 l M ) Cells were exposed to hyperosmotic stress for 1 h followed by 0 or 3 h incubation in sorbitol-free medium.

result in  50% hDlg-transfected HeLa cells, which is

a level tolerated by the cells Using these cells, we

examined the effect of hDlg mutation on the

progres-sion of both cell detachment and apoptosis induced by

sorbitol As shown in Fig 8, none of these processes

was significantly affected by the mutation D747A of

hDlg, indicating that its proteolysis is not sufficient for

the regulation of early apoptotic events

Discussion

Our aim was to gain a better understanding of the role

of hDlg phosphorylation in the adaptation of cells to

stress, by determining whether stress could regulate

hDlg degradation in cells exposed to changes in

envi-ronmental osmolarity We analysed the extent of hDlg

phosphorylation and hDlg protein levels in subconflu-ent and conflusubconflu-ent cells exposed to hyperosmolarity Our results show that hDlg is phosphorylated in response to osmotic stress, although the degree of hDlg phosphorylation upon sorbitol treatment is

 40% higher in subconfluent cells than in confluent cells This is probably because of the difference in hDlg localization observed In confluent cells, hDlg is present mainly in the membrane, at sites of cell–cell contact However, in subconfluent cells, hDlg localiza-tion is identical to its physiological kinase p38c, and both are localized diffusely throughout the cytosol, nucleus and membrane [3] Under these conditions, hDlg may be more accessible to the kinase and this may facilitate its phosphorylation after sorbitol treat-ment In addition, we have also shown that hDlg is largely degraded in cells exposed to hypertonicity and that this is not dependent on cell density

As mentioned previously, our objective was to inves-tigate in more depth the role of phosphorylation in hDlg degradation, but the results obtained from exper-iments using hDlg mutated at different p38c phosphor-ylation sites or using several kinase inhibitors, including the general kinase inhibitor staurosporine, demonstrated that under these conditions hDlg degra-dation was not regulated by its phosphorylation state

In addition, we showed that hDlg degradation is blocked by z-VAD, a general caspase inhibitor, and this indicates that hyperosmotic shock-induced loss of hDlg is mediated by caspases during apoptosis The caspase-dependent cleavage of many key structural and regulatory proteins contributes to the typical morphological changes, including the dismantling of cell–cell contact, seen during apoptosis As mentioned previously, hDlg is a scaffold protein, which has been implicated in the maintenance of cell polarity and cell adhesion [4] We report that hDlg is proteolysed by caspases during the apoptosis of HeLa (Fig S3) trig-gered by hyperosmolarity and also in CaCo-2 cells and mouse embryonic fibroblasts However, the mechanism

by which there is a different degree of apoptosis, and therefore of hDlg degradation, in HeLa cells than in HEK293 cells is unknown We speculate that in HeLa cells prolonged exposure to hypertonicity induces apoptosis because of a lack of or the dysfunction of the component(s) needed for the adaptive response of these cells to stress For example, it has been described that, in some cell types, the lack of the restoration of cell volume after cell shrinkage is associated with the concomitant appearance of apoptosis [23] However, the difference in the degree of apoptosis observed between HeLa and HEK293 cells may be due to a dif-ference in the sensitivity of these cells to the strength

0

5

10

15

Sorbitol Control

hDlg–D747A hDlg–D397A hDlg–WT None

B

3

4

5

6

7

Time after sorbitol release (h)

hDlg–D747A hDlg–D397A hDlg–WT None

A

Fig 8 Quantification of cells attached and apoptotic cells after

sor-bitol treatment (A) Cell attachment and (B) caspase 3 activation

were measured as indicated in Materials and methods, at the times

indicated – (A) 0 and 6 h or (B) 0 and 2 h – in nontransfected and

hDlg wild-type or hDlg–D397A or hDlg–D747A transfected HeLa

cells All other experimental details are described in the text Values

are means (± SE) of three independent experiments performed in

triplicate.