Tumor necrosis factor alpha sensitizes primary murine hepatocytes to Fas/CD95-mediated apoptosis in a Bim and Bid-dependent manner

Tumor Necrosis Factor ~ Sensitizes Primary Murine Hepatocytes to FaslCD95-lnduced Apoptosis in a

Bim- and Bid-Dependent Manner

Katiuin Schmich,

I

Rebekka Schlatter/* Nadia Corazza,'* Karine Sa Ferreira,4,5 Michael Ederer,2 Thomas Brunner,3 , 6 Christoph Borner,4.7, H and Irmgard Merfort

^l

Fas/CD95 is a critical mediator of cell death in many chronic and acute liver diseases and induces apoptosis in primary hepatocytes in vitro. In contrast, the pro inflammatory cytokine tumor necrosis factor IX (TNFIX) fails to provoke cell death in isolated hepato- cytes but has been implicated in hepatocyte apoptosis during liver diseases associated with chronic inflammation. Here we report that TNFIX sensitizes primary murine hepato- cytes cultured on collagen to Fas ligand (FasL)-induced apoptosis. This synergism is time-dependent and is specifically mediated by TNFIX. Fas itself is essential for the sensi- tization, but neither Fas up-regulation nor endogenous FasL is responsible for this effect.

Although FasL is shown to induce Bid-independent apoptosis in hepatocytes cultured on collagen, the sensitizing effect of TNFIX is clearly dependent on Bid. Moreover, both c- Jun N-terminal kinase ONK) activation and Bim, another B cell lymphoma 2 homology domain 3 (BH3)-only protein, are crucial mediators of TN FIX-induced apoptosis sensiti- zation. Bim and Bid activate the mitochondrial amplification loop and induce cyto- chrome c release, a hallmark of type II apoptosis. The mechanism of TNFIX-induced sensitization is supported by a mathematical model that correctly reproduces the biolog- ical findings. Finally, our results are physiologically relevant because TNFIX also sensi- tizes to agonistic anti-Fas-induced liver damage. Conclusion: Our data suggest that TNFIX can cooperate with FasL to induce hepatocyte apoptosis by activating the BH3- only proteins Bim and Bid. (HEPATOLOGY 2011;53:282-292)

E

nhanced apoptosis is critically involved in many acute and chronic liver diseases, and hepatocytes are the main cell type undergoing massive cell death during liver injUiy. This process is regulated by a

complex network of soluble and cell-associated apoptotic and inflammatory signals. I It is therefore increasingly im- portant to obtain insight into the mechanistic interplay of these signals to define new therapeutic strategies. In

Abbreviations: AuD, actinomycin D; AS?: aspartate aminotrtlnsfertlse; Bak, B cell lymphoma 2 homologous fmtagonist/killer; Bax, B cell lymphoma 2-tlSsociated X protein; Bcl2, B cell lymphoma 2; BI-/3, B cell lymphoma 2 homology domain 3; c-FLIR cellula.r Fas-associffting protein with death domain-like interleukin-1 beta-converting enzyme (FLlCE) i"hibitOlY protein; clAR cellular inhibitor of apoptosis; Diablo, diablo homolog; DISC, death-inducing signaling complex;

ELISA, enzyme-linked iml1lllnosorbent assay; FADD, his-associated death domain; FasL, Fas ligand; lOBS, fltal bovine serum; JNK, c-Jun N-terminrd kinase; KO, knockout; mAb, monoclollal ftntibocly; MOM!? mitochondrial membrtlne permeabiliZfltion; mRNA, messenger RNA; MTT; 3-(4,5-dimethylthiflzol-2-yl}-2,5- diphenyltetrtlzolilll1l bromide; N2A, neuroblastoma 2A; NF-KB, nuclear jactor kappa B; P-jNK, phosphOlylated c-jun N-terminal kinase; pBim, phoJphOlyiflted Bim; q1?7:PCR, qUftntitfltive retd-time polymerase chilin reaction; siBim, small intelfiring RNA tfllgeting Bim; siRNA, small illtt/firing RNA; Smac, second mitochondria-derived activator of caspases; 51'600125, anthlYl{1-9-cd/pyrazol-6{2H)-one; tBid, truncated Bid; TNF, tumor necrosis foctol; TNFR, tllmor necrosis foctor receptor; WT, wild type; )(JAR X-linked inhibitor ofapoptosis protein.

From the I Department of Pharmaceutical Biology and Biotechnolog)\ Albert-Ludwigs-University of fi'eiblllg, Freiburg, Germ({ny, llnstitute for System Dyn({mics, University ofStuttg({rt, Stuttgart, Germany; 'Biochemical Pharm({cology, Department of Biology, University ofKonstanz, Konstanz, Germany; "Institute of Mole cilIaI' Medicine and Cell Rese({rch, Albert-Ludwigs-University of Freibulg, Freibulg, Germany, 5 Faculty of Biology, Albert-Ludwigs-University of Freib'''g, Freiblllg, Germany, ('Division of Imnllmopathology, Imtitllte of Pathology, UnivelJity of Bern, Bern, Switzerland; ;Spemrtrlll Gmdllate School of Biology and Medicine, and and

"Cellter for Biological Signaling Studies, Albert-Ludwigs-University of Freiblllg, Freibltlg, Germ({ny.

This work was SIIpported by the Fedent! Ministly of Educ({tion fmd ResCflrch through a resetlrch gmnt (0313074D) Fom Hep({toSys (to Kathrin Schmich, Rebekkrr Schlatter, IV,rine Sfi Ferreim, Mich({el Edem; Christoph BornCl; fmd Irmgard MCljnrt). This work IVIIS also supported by the Spemfll1n Gradllate School of Biology and Medicine (to Christoph BorrlCl) ({nd by the Bioss Excellence Cluster (to Christoph BornCl); both {fre programs of the Excellence Initiative fmded by the German Research Foundation. Kllrine Sa' Ferreira is supported hy the German Reseanh Foundation (GRKII04), fmd Thomas Brunner and Nadi({ CoraZZfl are supported by the Swiss National Science Foundation (310000-120427 and 3/0000-121854).

'These flutholJ contributed equally to this work.

282

http://dx.doi.org/10.1002/hep.23987

Konstanzer Online-Publikations-System (KOPS) URN: http://nbn-resolving.de/urn:nbn:de:bsz:352-142378

the liver, apoptosis is mainly initiated by the death recep- tor ligands Fas ligand (FasL; CD95L) and tumor necrosis factor a (TNFa).2

After ligand binding, death receptors recruit the adaptor Fas-associated death domain (FADD) and pro- caspase-8 to their intracellular face, and this forms the death-inducing signaling complex (DISC).3 By this as- sembly, procaspase-8 is autoprocessed and activated, and it can then trigger two different apoptotic signal- ing pathways. In so-called rype I cells, such as lympho- cytes, active caspase-8 directly cleaves and activates procaspase-3 to induce efficient cell death execution."

In rype II cells, such as hepatocytes, apoptosis induc- tion first requires caspase-8-mediated cleavage of BH3-inreracting domain death agonist (Bid) into its truncated form [truncated Bid (tBid)]. tBid belongs to the subclass of B cell lymphoma 2 homology domain 3 (BH3)-only B cell lymphoma 2 (Bcl2) family mem- bers (e.g., Bcl2-like protein 11 [Bim], p53-up-regu- lated modulator of apoptosis (Puma), and Noxa), which sense apoptotic stimuli and convey the death signals to B cell lymphoma 2-associated X protein (Bax) and B cell lymphoma 2 homologous antagonist/

killer (Bak) activation on mitochondria. Although it is still unclear how this activation occurs,s it has become well accepted that Bax and Bale are essential for mito- chondrial membrane permeabilization (MOMP) and the release of apoptogenic factors such as cytochrome c and second mitochondria-derived activator of caspases (Smac)/diablo homolog (Diablo).6 Although cyto- chrome c activates the apoptotic peptidase activating factor I (APAF-l )/caspase-9 apoptosome, which results in effector caspase-3/caspase-7 activation, SmaclDiablo neutralizes the caspase-9 and caspase-3 inhibitor X- linked inhibitor of apoptosis protein (XIAP). Recently, XIAP has been shown to determine the rype IIIl FasL signaling switch in hepatocytes and [3-pancreatic celll because a large abundance of XIAP requires neutraliza- tion of its caspase-3-inhibiting activiry by rype II sig- naling to allow effective cell death. S.R

FasLlCD95L and its corresponding receptor Fas/

CD95 play pivotal roles in the immune system; they induce the death of infected cells and obsolete lympho- cytes and thereby protect against autoimmuniry and tu- mor development.⁴.? Furthermore, Fas is constitutively

283

expressed on the surface of hepatocytes and is impor- tant to hepatic health and disease. Mice treated with a lethal dose of agonistic anti-Fas antibody die because of massive hepatocyte apoptosis and liver failure.^lo This cell death is dependent on Bid because Bid-deficient mice are resistant to Fas-induced hepatocellular apopto- sis, fulminant hepatitis, and subsequent liver failure. ^{I I} These findings indicate that in vivo hepatocytes die in response to FasL via the rype

n

^{signaling pathway?}

However, we have shown recently that isolated primary hepatocytes cultured on collagen change their apoptosis signaling from rype II to the Bid-independent type I pathway,12 and this suggests that the rype JIll decision depends not only on the expression of endogenous pro- teins, such as XIAp, but also on external factors.

TNFa is a pleiotropic cytokine that induces a vari- ery of cellular responses, such as inflammation and cell proliferation, mainly through activation of the nuclear factor kappa B (NF-ICB) signaling cascade. Unlike FasL, the association of TNFa with its main receptor tumor necrosis factor receptor I (TNFRl) does not primarily lead to cell death in most cell rypes, includ- ing hepatocytes.13 After activation of TNFRl, mem- brane-bound complex I is first formed and rapidly activates survival transcription factor NF_KB.14 To sig- nal for cell death, a second complex, receptor-free complex II, has to assemble in the cytoplasm and recruits FADD and caspase-8 to activate caspase-3/cas- pase-7.11 Under normal conditions, complex II forma- tion is blocked by cellular Fas-associating protein with death domain-like interleukjn-l beta-converting enzyme (FLICE) inhibitory protein (c-FLIP) and NF-ICB sur-

. al ' I' 1)1(, H h' ul' b '

VIV signa mg. ' owever, t IS reg anon can e cir- cumvented by yet another TNFa-activated apoptotic signaling pathway that involves activation of c-Jun N- terminal kinase ONK). It has been shown that JNK mediates TNFa-induced apoptotic signaling by the phosphorylation and activation of the BH3-only protein Bim. U.17 In agreement with this notion, TNFo:-induced hepatocyte apoptosis has recently been reported to require both Bim and Bid in vivo. I R In this study, active caspase-8 generated tBid, whereas active JNK phospho- rylated and stabilized Bim, and the interplay of both processes was necessary to induce full Bax/Bak activa- tion and hepatocyte apoptosis in response to TNFa.

Address reprint requests to: Irmgftrti Me/jim. Ph.D .. Department o/Phflrmacellticftll3ioiogy flntll3ioteclmoiogy, Aibert-Llldwigs-Univem'ty 0/ Freib'.tlg, Stefi/ll- Mcier-Stmle 19. 79104 Freiblllg. Germany. E-mflii: irmgflld.mcljim@p/Jflrlllazic.uni-ji-cib/llg.de; fox: +497612038383.

PotCllt;fti conflict of interest: Dr. Brunner owns stock ill Novtlrtis flnd Roche.

Here we show that a similar Bim/Bid interplay is used by TNFa to sensitize primary mouse hepatocytes to FasL-induced apoptosis in vitro. We also demon- strate this sensitizing effect toward anti-Fas-induced liver damage in vivo. Although TNFa itself is nona- poptotic, it markedly enhances FasL-induced hepato- cyte apoptosis via both the JNK/Bim and Bid signal- ing pathways. These data confirm that TNFa is capable not only of engaging the JNK/Bim apoptotic pathway but also of restoring type II signaling on col- lagen-cultured primary hepatocytes. This crosstalk is supported by a systems biology approach because we present a qualitative mathematical model that correctly reproduces the biological findings.

Materials and Methods

Isolation and Cultivation of Primary Mouse Hep- atocytes. Primary hepatocytes were isolated from 8- to l2-week-old wild-type (WT), Bid-^I- , XlAP-¹- , Fas-^I-,

or FasLgld/gld C57BLl6 mice with the collagenase perfusion technique (see the supporting information for details).

Induction of Hepatitis and Histology. Young, adult WT C57BLl6 mice were injected intravenously with TNFa (40 ftg/kg of body weight; Peprotech), and this was followed by an intravenous injection with an anti-Fas antibody (clone J02; BD Bioscience-Pharmin- gen) at a dose of 80 ltg/kg of body weight 2 hours later. Liver damage was assessed 5 hours later by the measurement of the serum aspartate aminotransferase (AST) levels with a commercially available kit (505-0r, Teco Diagnostics). Five-micrometer liver tissue sections were stained with hematoxylin and eosin for histological assessment. All animals were handled and housed under specific pathogen-free conditions, and animal experi- ments were reviewed and approved by the animal exper- imentation review board of the Stare of Bern.

RNA Interference. Smart pools of mouse Bim small interfering RNA (siRNA) duplexes and nontargeting control duplexes were purchased from Dharmacon (ON-TARGETplus SMARTpool); the Lipofectamine RNAiMAX transfection reagent was obtained from Invitrogen. For siRNA transfection, cells were reverse- transfected with 10 nM siRNA with Lipofectamine in the Opti-MEM medium according to the manufac- turer's instructions. Effective knockdown was verified by quantitative real-time polymerase chain reaction (qRT-PCR) and immunoblotting after different times (Supporting Fig. 1).

Other Experimental Procedures. Other experimen- tal procedures are described in detail in the supporting information. These include the mice, preparation of

total, cytosolic, und mitochondrial lysates, western blotting, quantification of neuroblastoma 2A (N2A) FasL, quantification of Vl q TNFa-neutralizing anti- body,

o

EVDase, also known as caspase-3 like assay, assay, 3-( 4,5-dimethylthiazol-2-yl) -2,5-di phenyl tetrazo- lium bromide (MIT) viability assay, Cell Death Detec- tion enzyme-linked immunosorbent assay (ELISA), RNA isolation, complementary DNA synthesis and qRT-PCR, and cytochrome c ELISA.

Results

TNFrJ. Preincubation of Primary Mouse Hepatocytes Increases FasL-Induced Caspase-3 Activation and Cell Death in a

Time-Dependent Manner

We previously reported that FasL induces the apo- ptosis of collagen-cultured primary murine hepatocytes via the type I signaling pathway, but only to a moder- ate extent. 12 In this study, we focused on the crosstalk of FasL with the proinflammatory cytokine TNFex. We preincubated collagen-cultured primary murine hepa- tocytes with 25 ng/mL TNFa for 12 hours, and this was followed by a treatment with 50 ng/mL FasL for 6 hours. As expected, untreated and TNFa-treated he- patocytes showed a typical binuclear morphology and no signs of cell death over an incubation period of 18 hours (Fig. 1A). In contrast, as previously reported, cells treated with FasL for 6 hours showed hallmarks of apoptosis such as cell shrinkage and plasma mem- brane blebbing.¹² When the cells were preincubated with TNFa for 12 hours before the FasL treatment, they underwent a significantly higher degree of apo- ptosis (Fig. lA). These findings could be confirmed by the measurement of the effector caspase-3/caspase-7 ac- tivity in response to the different treatments. As shown in Fig. lB, the longer the hepatocytes were cultured (J 2, 24, or 48 hours), the more caspase-3/caspase-7 ac- tivity they displayed with a 6-hour FasL treatment. If during this culturing the cells were exposed to TNFa, the caspase-3/caspase-7 activities further increased and were consistently higher than those with FasL alone.

Importantly, a minimum preincubation time of-approxi- mately 2.5 to 3 hours was needed for TNFa to exert its sensitization on FasL-induced caspase-3/caspase-7 activa- tion, and this indicated that the TN Fa effect was not immediate (Fig. 1 C). We also tested the dose depend- ence of the sensitization and found that varying the TNFa concentrations from 10 to 50 ng/mL did not modulate the preincubation time required for sensitiza- tion (Supporting Fig. 2). Moreover, the sensitization was clearly caspase-dependent because cell death (as

A

phase contrast

B

⁸⁰

60

~ ⁴⁰

20

0

- ^o

^{~ control TNF FasL}

_ TNF+FasL

12+6 24+6 48+6

time [h]

285

C

^{40 _ control}

~ FasL

30 D TNFa

• TNFa+FasL

~ ²⁰

10

0

6 0.2+6 0.5+6 1 +6 2+6 3+6 4+6 12+6 time [h]

Fig. 1. TNFo. sensitizes primary hepatocytes to FasL-induced caspase-3/caspase-7 activation and apoptosis in a time-dependent manner. (A) Phase contrast microscopy of primary murine hepatocytes that were untreated (control) or treated with TNFo. (25 ngjmL), N2A FasL (50 ngjmL), or a combination of the two for the indicated times (original magnification, 200x). (B) Caspase-3/caspase-7 activity determined by a fluoro- genic DEVDase assay of cells treated with TNF~ FasL, or TNFo. and FasL after the indicated times. FasL was added during the last 6 hours of the total treatment time (18, 30, and 54 hours, respectively). (C) Kinetics of caspase-3/caspase-7 activation of cells preincubated for different times with TNFo. (FasL treatment for 6 hours). Values represent means of at least three independent experiments and standard deviations. *P <

0.001 versus FasL-treated cells and #P < 0.05 versus FasL-treated cells (Student t test).

measured by the MTT assay) was effectively blocked in the presence of a pancaspase inhibitor (quinoline-Val- Asp O-phenoxy, non-O-methlyated (q VD-OPh); Sup-

porting Fig. 3).

Sensitization of FasL-Induced Apoptosis by TNF(X Is Specific for TNFrr. and the Opposite Sensitization Cannot Be Observed. Because other factors in the growth medium may modulate FasL-induced apoptosis signaling, we first confirmed that the sensitizing effect is specifically mediated by TNFcx. We therefore added TNFIX-neutralizing antibodies produced by the VI q hybridoma cell line (J 00 J1L of the culture superna- tant) to the primary hepatocytes 30 minutes before TNFIX and FasL stimulation. TNFIX-neutralizing anti- bodies effectively prevented the sensitization because caspase-3/caspase-7 activity did not increase beyond that measured with FasL alone (Fig. 2A). We then tested the inverse scenario (i.e., whether FasL was also able to sensitize hepatocytes to TNFIX-induced apopto- sis). For that purpose, cells were first treated with FasL, and 2 hours later, TNFIX was added for a total of 4 hours before the measurement of active caspase-31 caspase-7. As demonstrated in Fig. 2B, FasL-induced caspase-3/caspase-7 activity could not be further increased by TNFcx. This finding confirms that the ap-

optosis sensitization effect of TNFIX is specific for this cytokine, needs a certain time threshold (as shown in Fig. 1 C), and involves a molecular mechanism that cannot be engaged by FasL. To completely exclude the implication of growth factors, we tested the role of fe- tal bovine serum (FBS) in the sensitization effect. As shown in Supporting Fig. 4, FBS neither enhanced nor inhibited the sensitization of FasL-induced apoptosis by TNF~ but primary hepatocytes turned out to be more sensitive toward FasL-induced apoptosis in the presence of FBS (see also Walter et aI.¹²).

TNF(X Sensitization Is Not Mediated via Tran- scriptional Up-Regulation of Fas ^01' FasL but Inter- feres With the Fas Signaling Pathway. To uncover

the molecular mechanism of the TNFIX sensitization, we tested various possibilities for TN FIX crosstalk with the Fas/FasL system. First, we compared apoptosis between WT and Fas 1- hepatocytes to investigate the role of Fas. As shown in Fig. 3A, Fas-^I- hepato- cytes did not show any caspase-3/caspase-7 activation in response to FasL or sensitization by TNFcx. In con- trast, caspase-3/caspase-7 activity levels were unchanged between WT and Fas-^I- cells when they were treated with TNFlX/actinomycin 0 (ActO), and this indicated that TNFIX-mediated sensitization to FasL-induced

_ WT

o

TNFmAb

Fig. 2. The sensitizing effect is specifically mediated by TNFa. (A) Caspase-3/caspase-7 activity of cells treated with TNFCI, FasL, or a combination of the two with or without the TNFa.-neutralizing antibody.

ffP < 0.05 versus FasL-treated cells (Student t test). (8) Cells treated

in the opposite order (first with FasL for 2 hours and then with TNFa.

for another 4 hours versus FasL or TNFo: alone for 6 hours). Values represent means of at least three independent experiments and stand- ard deviations. Abbreviation: mAb, monoclonal antibody.

apoptosis required Fas. Therefore, we next tested whether sensitization could be due to up-regulation of endogenous Fas by TNFct However, the qRT-PCR analysis did not reveal any induction of Fas messenger RNA (mRNA) in response to TNFC( (data not shown). Besides Fas, TNFC( could up-regulate endogenous FasL and thereby amplifY the FasL-induced apoptotic response. To test this hypothesis, we analyzed TNFC( sensitization in F _'as L^gld/"_" ^ld h epatocytes, w h' IC 1 I express a mutant rorm C 0 f FasL that cannot bind Fas. As shown in Fig. 3B, the loss of endogenous FasL production did not signifi- can t1y reduce the enhanced caspase-3/ caspase-7 activa- tion because of TNFC( preincubation of the FasL-treated cells. These findings indicate that TNFC( impinges on the intracellular FasL signaling pathway rather than the regulation of Fas or FasL in order to sensitize primary hepatocytes to FasL-induced apoptosis.

Bid but Not XIAP Is Critical for TNFrx Sensitiza- tion. Because our findings so far suggested direct crosstalk between TNFC( and Fas signaling, we per-

formed a detailed analysis of the components of the twO signaling pathways. We recently reported that FasL-induced apoptosis of collagen-cultured primary

. mouse hepatocytes occurred independently of Bid.

This was in contrast to apoptosis induced by TNFC(/

ActO, which still required Bid (type II signaling). 12

We therefore tested whether this was also the case for the sensitization effect of TNFC( on FasL-induced apo- ptosis. Indeed, although Bid-^I- hepatocytes showed the same caspase-3/caspase-7 activation in response to FasL that WT cells showed, the increased caspase-3/caspase-7 activation due to treatment with TNFC( and FasL was entirely abolished (Fig. 4A). Both cell death (based on the MIT assay; Supporting Fig. 5A) and apoptosis-asso- ciated DNA fragmentation (Supporting Fig. 5B) were reduced in Bid-^I- hepatocytes versus WT cells when they were treated with TNFC( and FasL, and this sup- ported the caspase data. Additionally, Bid was processed into its active form (tBid) in cells treated with TNFct and FasL, whereas TNFC( alone did not lead to any tBid formation (Fig. 4B). However, TNFct induced

A

control TNF FasL TNF+FasL TNF+ActD

_ WT

D FasLgldlgld #

control TNF FasL TNF+FasL

Fig. 3. Fas (but not newly synthesized FasL) is essential for TNFo:- mediated sensitization. Caspase-3/caspase-7 activity assays of (A) 'NT and Fas / hepatocytes and (8) 'NT and FasLgld/gld hepatocytes treated with TNFo: (25 ng!mL), FasL (50 ng!mL), or a combination of the two for 12 and 6 hours. Values represent means of at least three inde- pendent experiments and standard deviations. 'P < 0.001 versus FasL- treated cells and ffP < 0.05 versus FasL-treated cells (Student t test).

B

C

I

Q)

t;

.S:

g

'0

Z « ex: E

'0

00

_ WT

12+6 12+6 12+6 24+6 24+6 24+6 TNFa 25ng/mL[hj FasL 50 ng/mLl6hj SP600125 25 ~M

10

8

6

4

2

'l- time [hJ

-+- TNFa

Fig. 4. Loss of Bid abrogates the sensitizing effect, and Bid is pro- cessed to its active form tBid upon treatment with TNFCI. and FasL. (A) easpase-3/caspase-7 activity of WT and Bid-/- hepatocytes treated with TNFCI. (25 ngjmL). FasL (50 ngjmL), or a combination of the two for 12 and 6 hours. Means of at least three independent experiments and standard deviations are shown. 'P < 0.001 versus FasL-treated cells (Student t test). (B) Anti-tBid/Bid western blot analysis of total extracts of cells treated with TNFCI. and/or FasL for the indicated times or additionally preincubated with the JNK inhibitor SP600125 (20 pM). Actin is shown as the loading control. (e) Time course analysis of Bid mRNA expression in response to TNFCI. (25 ngjmL) as meas- ured by qRT-peR (TaqMan). Values represent means of at least three independent experiments and standard deviations.

increased expression of the Bid protein (Fig. 4B) and mRNA (Fig. 4C) and this further strengthened a crucial role of this protein in the sensitizing mechanism. Bid processing was also observed with FasL alone, but this did not contribute to apoptosis induction (Fig. 4A).

Thus, our results show that although Bid is not required for FasL-induced apoptosis on collagen-cul-

287

tured hepatocytes, it is absolutely crucial for the TNFa sensitization of this process.

XlAP is an endogenous inhibitor of caspase-9 and effector caspase-3/caspase-7 and restrains apoptosis along the type I pathway unless it is neutralized by apoptogenic factors emanating from mitochondria.

Accordingly, as we previously showed, XlAP-/- hepa- tocytes exhibited 10-fold higher caspase-3/caspase-7 activity in response to FasL than WT cells (Supporting Fig. 6). This activity was not further increased by TNFa preincubation. However, a slight sensitization was seen at low FasL doses (10-20 ng/mL). This indi- cates that deletion of XlAP does not abrogate the TNFa sensitization to FasL-induced apoptosis. Impor- tantly, XlAP protein (Supporting Fig. 7) and mRNA (Supporting Fig. 16C) remained nearly unchanged during TNFa preincubation. Thus, XlAP turned out to be dispensable for the sensitizing effect of TNFa.

TNF(J. Activates JNK, and JNK Inhibition Blocks Apoptosis Sensitization

by

TNFa. Activation of JNK has been implicated in TNFa-induced apoptosis in several cell types, including hepatocytes. 19,20 We there- fore monitored the active phosphorylated form of JNK by anti-phospho-JNK western blot analysis in primary mouse hepatocytes treated with TNFa. TNFa/ActD, which is known to induce apoptosis by prolonged JNK activation,21 was included as a positive control. We found that TNFa induced early phosphorylation of JNK in the first 30 minutes, although this was not as

high as that with TNFa/ ActD after 6 or 8 hours (Fig.

5A). To investigate the significance of this early JNK activation for apoptosis sensitization, we preincubated primary hepatocytes with the JNK inhibitor anthra[1-9- cdJpyrazol-6(2H)-one (SP600125; 25{IM), which was followed by FasL or a consecutive treatment with TNFa and FasL. Strikingly, JNK inhibition could effectively block the sensitizing effect of TNFa on caspase-3/cas- pase-7 activation because DEVDase activity levels in the presence of SP600 125, TNFa, and FasL were essentially the same as those with FasL alone (Fig, 5B), This decrease in caspase-3/caspase-7 activity resulted in a sig- nificant reduction in actual cell death and apoptosis (Supporting Fig, 8), and this supported the role of JNK in the sensitization, In contrast, the p38 mitogen-acti- vated protein kinase inhibitor RN3503 (10 {1M) had no effect (Supporting Fig, 9), and this indicated that JNK (but not p38 mitogen-activated protein kinase) was cru- cially involved in apoptosis sensitization by TNFa.

Bim Is Essential for the Sensitizing Effect of TNFa. It has recently been reported that Bid and Bim are both essential for TNFa-mediated hepatocyte apopto- sis in vivo. I 8 Furthermore, it is known that the

10 20 30 3SO 3SO 480 480 TNFa 25 ng/mL[minl ActO 0.4 ~glmL {hi

~----~~~--~~--~~~~----. +-p54 P·JNK

~;=~i=:::::;==========~

^{+- p46}

#,'

JNK

_ control

CJ TNF

~ FasL _ TNF+FasL

Fig, 5, TNFIX activates JNK, and both JNK and its downstream target Bim are crucial for the sensitization, (A) Western blot analysis of P- JNK after treatment with TNFIX alone or ActD for the indicated times, (B) easpase-3/caspase-7 activity in response to TNFIX and FasL with or without pretreatment with the JNK inhibitor SP600125 (25 11M) for 12 and 6 hours, Means of at least three independent experiments and standard deviations are shown, 'P < 0,001 versus FasL-treated cells (Student t test), (e) easpase-3/caspase-7 activity of primary mouse hepatocytes not transfected or transfected with control siRNA or Bim siRNA (10 nM) after treatment with TNFIX (25 ngfmL), FasL (50 ngfmL), or both for 12 and 6 hours, Values represent means of at least three independent experiments and standard deviations, 'P

< 0,001 versus FasL-treated cells, " P < 0,01 versus FasL-treated

cells, and ffP < 0,05 versus Bim siRNA-transfected cells treated with TNFIX and FasL (Student t test), Abbreviation: P-JNK, phosphorylated c-Jun N-terminal kinase,

proapoptotic activity of Bim can be regulated by JNK- mediated phosphorylation, J 7,22 Consequently, we stud- ied the role of Bim in the TNFa sensitization mecha- nism by down-regulating Bim expression by siRNA.

The Bim mRNA and protein were effectively down-

regulated by small interfering RNA targeting Bim (siBim); this was verified by qRT-PCR (Supporting Fig, lA) and western blot analysis (Supporting Fig.

1 B), respectively, Strikingly, although control siRNA did not affect caspase-3/caspase-7 activity levels in cells treated with TNFa and FasL, siBim significantly reduced them to the levels measured with FasL alone (Fig. 5C). In addition, the loss of Bim resulted in decreased apoptosis"associated DNA fragmentation and cytotoxicity upon treatment with TNFa and FasL (Supporting Fig, 10). Thus, both Bid and Bim seem to be required for the sensitization effect of TNFa on the FasL-induced apoptosis of primary mouse hepato- cytes. Because JNK is also crucial for this effect and the inhibition of JNK could not abrogate tBid forma- tion (Fig. 4B), we suggest that the implication of Bim involves its JNK-mediated phosphorylation, as previ-

17"2 ':3

ously shown. ,- ,-

TNFrx Sensitization Involves Restoration of TYpe II Signaling in Collagen-Cultured Hepatocytes. Both Bid and Bim relay apoptotic signals to the activation of Bax/Bak, which in turn triggers MOMP and the release of cytochrome c and other apoptogenic fac- tors (type II signaling). We therefore tested whether TNFa-mediated sensitization to FasL-induced apoptosis involved cytochrome c release. For that purpose, we pre- pared cytosolic and mitochondrial fractions from TNFa- treated, FasL-treated, or TNFalFasL-treated hepatocytes, verified their purity by western blot analysis (Supporting Fig. 11), and determined the concentration of cytosolic

:::::J 0, E

150

oS 100

l.l Ql

E o

.c

50 u o

%

_ WT

o

Bid·^l-

Fig, 6, TNFIX sensitization induces JNK-dependent and Bid-depend- ent cytochrome c release, The cytochrome c concentration was deter- mined by ELISA in the cytosol of WT and Bid-^I- hepatocytes after treatment with TNFIX (25 ngfmL), FasL (50 ngfmL), or both with or without pretreatment with the JNK inhibitor SP600125 (25 11M) for 12 and 6 hours, Means of at least three independent experiments and standard deviations are shown, 'P < 0,001 versus FasL-treated cells and "P < 0,01 versus cells treated with SP600125, TNFCI, and FasL.

Fig. 7. TNFa preincubation sensitizes mice to anti-Fas-induced liver damage.

(A) Liver aminotransferase (AST) serum levels in mice treated with J02 (agonistic anti-Fas antibody; 80 fig/kg of body weight) alone for 5 hours or pretreated for 2 hours with TNFa (40 fig/kg of body weight) before J02 administration. Values represent means and standard deviations (four mice per group). One of two repre- sentative experiments is shown. (8) His- tological analysis of liver sections from mice injected with J02 alone or injected with TNFa 2 hours before J02 administra- tion. Low-power and high-power magnifi- cations of representative samples are shown. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

B

Jo2

TNFa+Jo2

A

cytochrome c by ELISA. As shown in Fig. 6, cytochrome c could indeed be detected in the cytosol of hepatocytes treated with TNFIX and FasL, whereas neither TNFIX nor FasL alone promoted any cytochrome c release, as previ- ously described.¹² Importantly, cytochrome c release did not occur in TNFIX/FasL-treated Bid-^I- hepatocytes or when JNK was inhibited (Fig. 6), and this supported the notion that Bid and JNK were involved in the sensitiza- tion mechanism. These results indicate that TNFIX enhances FasL-induced apoptosis of collagen-CLutured pri- mary hepatocytes by activating a Bid-dependent and Bim-dependent type II apoptosis pathway. We also inves- tigated whether antiapoptotic Bel2 family members were modulated during TNFIX sensitization, but neither B cell lymphoma extra large (Bel-XL) nor myeloid cell leukemia sequence 1 (Mel-I) levels were up-regulated or down- regulated (Supporting Fig. 12).

TNF(I. Sensitizes Mice In Vitro to Anti-Fas- Induced Liver Damage. To test whether the sensitiz- ing effect in cultured hepatocytes could also be

289

Jo2 TNFa+Jo2

20x 100x

observed in vivo, mice were injected with recombinant murine TNFIX followed by anti-Fas antibody 002), and liver damage was assessed by the measurement of AST levels. Strikingly, these first experiments revealed an increase in AST levels (Fig. 7 A) and tissue damage, which was shown by an enhancement of apoptotic cells (Fig. 7B) when mice were challenged with TNFIX and J02 versus J02 administration alone. Before final conclusions can be drawn, further experiments have to be performed. Nevertheless, these results indicate that the sensitizing effect reported here could be physiologi- cally and clinically relevant.

Mathematical Modeling Confirms the Mechanism of Sensitization by TNF(1. A qualitative mathematical model of the crosstalk between TNFIX and FasL signal- ing was built to further analyze the sensitizing mecha- nism. The model is based on ordinary differential equations using mass action kinetics, and its structure is illustrated in Fig. 8A. TNFIX and FasL are consid- ered possible model inputs that activate their respective

A

^{TNF FAS}

B

¹⁰⁰

80~---

... , ... , l:

⁶⁰

I T

_V

I '"

_Va

" ~ ~~=~~;;;;;;~

_{0 2 ~ 6 8}

v" time [h]

fpBlm: Bel2 fiBId:

0

¹⁰⁰

E

¹⁰⁰

14 16 18

time[h] 20

:.~' .. : : .. v'P.: 801+ ... - - - - -... , 80

l:~

l:oo

f u

⁴⁰ ~ fi ~

'" 20 '" 20

O~ O~2~4~6~8 ~ 1~ 0 ~ 12 ~ 14 ~ 16~18~2~0

time[h]

14 16 18 20

time [h]

B·O: pJNK[t} pElim[t] Bc/2[t] Ca'(t] tElid[t]

BaxBak'[t] CytC.,,[t] XIAP[t] C3'[t]

E: wild type XIAP KO Bid KO I FasL:

dashed, circle I TNF+FasL: Joined, star

Fig. 8. Mathematical modeling is able to reproduce the sensitizing effect. (A) Illustration of the model structure. (B·D) Simulation results for

wr hepatocytes after stimulation with (B) TNFCI, (C) FasL, or (D) both for 12 and 6 hours. Invisible curves are zero. (E) Simulation results for wr,

Bid I ,and XIAP I hepatocytes after stimulation with FasL or with TNFa and FasL for 12 and 6 hours versus the measurement results of this study. Abbreviations: KO, knockout; cytcfree, cytochrome c released to the cytosol. [Color figure can be viewed in the online issue, which is avail·

able at wileyonlinelibrary.com.]

pathways ~o converge on Bax/Bak activation. We assume that phosphorylated Bim (pBim) and tBid act similarly on Bax/Bak activation but with different pa- rameters (vG and vd. Both can also be neutralized by Bcl2 family members (Bcl2). In the model, the release of cytochrome c is realized via a step nmction triggering 100% release at a threshold of 90% Bax/Bak activation.

The model equations, parameter values, and sensitiviry analysis are provided in the supporting information.

Simulation results for WT hepatocytes after treat- ment with TNFa, FasL, or TNFa and FasL are shown in Fig. 8B-D. Analogous simulations are provided in the supporting information for Bid-I- and XlAP-I- cells (Supporting Figs. 13 and 14). In Fig. 8E, the simulation results for caspase-3 activation are com- pared to the respective measurements for WT and XlAP-I- and Bid-^I- hepatocytes. Overall, the model is able to accurately reproduce the observed sensitizing effect in all studied genorypes.

Discussion

TNFa is a proinflammatory cytokine that plays a crucial role in both liver regeneration²⁴ and liver cell apoptosis during disease states. I In this article, we report that TNFa sensitizes primary mouse hepato- cytes to FasL-induced apoptosis in a Bid-dependent and Bim-dependent manner. We further show that this crosstalk involves JNK activation and most likely Bim phosphorylation, cleavage of Bid, and, consequently, activation of the rype I1 mitochondrial pathway and

results in cytochrome c release and effector caspase-3/ caspase-7 activation. Controversial results have so far been reported concerning the crosstalk of TNFa and FasL in apoptosis induction. On the one hand, TN Fa has been shown to confer resistance to Fas-induced cell death in eosinophilic acute myeloid leukemia cells because of its NF-KB-mediated antiapoptotic func- tions.²⁵ In this respect, we analyzed some rypical anti a- poptotic NF-KB target genes such as cellular inhibitor of apoptosis 2 (cIAP2), c-FLIp, and XlAl~ but we found that they were only moderately up-regulated (if ever) in response to TNFa (see Supporting Fig. 16).

clAPl protein was not at all detected in hepatocytes (see also Walter et aI.¹²; data not shown). On the other hand, several studies have indicated that TNFa positively regulates Fas-mediated apoptosis. In one case, TNFa could even overcome the Fas resistance of human lung fibroblasts2G

by allowing more FADD adaptor to bind to Fas and therefore increase DISC formation and FasL- mediated apoptotic signaling. In contrast to human lung fibroblasts, primary mouse hepatocytes do not seem to have impaired DISC formation because they are quite sensitive to FasL-induced apoptosis.

To obtain evidence for the physiological relevance of TNFa/FasL crosstalk, Costelli et al.²⁷ used gene target- ing to show that a loss of TNFRI and TNFR2 pro- tects mice from anti-Fas antibody-induced liver injury.

Our results confirm these findings and demonstrate that TNFa is necessary for efficient FasL-mediated hepato- cyte apoptosis. However, the exact mechanism of the interplay of the two pathways was not unraveled in the

previous study. It was shown that liver tissue levels of Fas and FasL as well as Fas expression on the hepatocyte surface were unchanged, but Bel2 was up-regulated upon TNFRI and TNFR2 depletion; this indicates that TNFa may regulate Bel2 family members.27 This again is consistent with our finding that neither Fas up-regulation nor endogenous FasL is critical for the TNFa sensitizing effect, and changes in members of the Bel2 protein family could be the underlying mech- anisms for the involvement of the type II mitochon- drial pathway in the sensitization process.

On the other hand, it is widely accepted that TNFa fails to induce apoptosis in hepatocytes under normal conditions because of activation of the NF-/(B survival pathway. Inhibition of this pathway restores apoptosis, and one mechanism involves the inducement of sus- tained activation of JNK.^2L•28

This prolonged JNK activation has been shown to be crucial for TNFa- mediated hepatocyte apoptosis but not for Fas.²⁰ Our findings confirm that TNFa alone does not induce he- patocyte apoptosis but, under transcriptional arrest with ActD, leads to sustained JNK activation critical for apoptosis. Interestingly, TNFa also induces early transient JNK activation, which by itself does not directly induce apoptosis but is critical for TNFa- mediated sensitization to FasL-induced apoptosis. Sev- eral reports have indicated that JNK modulates the proapoptotic activity of the BH3-only protein Bim by phosphorylation.17.22,23 This specific phosphorylation causes either the release of Bim from its sequestration to the microtubular dynein motor complex or the sta- bilization of the Bim protein; both can induce Bax/

Bak-dependent apoptosis. However, regulatory phos- phorylation of Bim by other kinases such as extracellu- lar signal-regulated kinase can induce the opposite effect and lead to proteasomal degradation and protec- tion from apoptosis.29

Hence, the regulation and out- come of Bim phosphorylation have to be further elarified in hepatocytes through, for example, the identification of the exact phosphorylation sites and the expression of phosphorylation-defective Bim mutants. The role of JNK-mediated Bim phosphorylation in hepatocyte apo- ptosis has recently been substantiated in vivo.¹⁸ The authors showed that lipopolysaccharide/galactosamine- treated mice died because of TNFa-mediated fatal hepa- titis and demonstrated that this apoptosis was dependent on Bid and Bim. Bim was shown to be phosphorylated by JNK and, consequently, redistributed from microtu- buies to the cytosol; there, it induced apoptosis in coop- eration with caspase-8-cleaved tBid. Remarkably, only the loss of both Bid and Bim protected mice from lipo- polysaccharide/ galactosamine-induced hepati tis. Similar

291

findings have been observed for TNF-related apoptosis- inducing ligand (TRAIL), which enhances Fas-induced hepatocyte apoptosis and liver damage via activation of the JNK-Bim axis23

; this suggests some overlapping effects of different TNF family members. Our results with cultured primary mutine hepatocytes support the aforementioned mechanism. TNFa preincubation led to JNK activation, and the inhibition of JNK and the loss of Bim abolished the sensitizing effect; however, FasL- induced apoptosis remained unchanged. In addition, sen- sitization was mitigated by the loss of Bid. In our study, TNFa needs to crosstalk with Fas to exert its apoptosis- sensitizing effect. We recently reported the unexpected finding that in collagen-cultured primary mouse hepato- cytes, Fas signaling switches from a rype II, Bid-de- pendent apoptotic signaling pathway to a rype I, Bid- independent apoptotic signaling pathway. As shown here, TN Fa is obviously able to restore the rype II sig- naling pathway by a so far unknown mechanism. It will be crucial to identifY these crosstalk points between TNFa and FasL signaling. Our data suggest that Bim and Bid may be part of these points. Both act by triggering Bax/Bak-mediated MOMP and cyto- chrome c release, but perhaps this occurs efficiently only when both are indeed present and activated.

TNFa would activate Bim via JNK and regulate Bid in a so far unknown way such that it becomes required for FasL-induced apoptosis. This would explain why TNFa-induced sensitization is impeded in both Bim knockdown and Bim -/- hepatocytes. We therefore sug- gest that Bim and Bid can only cooperatively activate the mitochondrial amplification loop in hepatocytes and that this is crucial for the observed increased sensitiviry to FasL-induced apoptosis.

The presented mathematical model accurately repro- duces the sensitizing effect and will promote further directions for future research. Sensitivity analysis reveals the sensitizing mechanisms to be very robust, although the model contains only the most important players.

Most critical interactions for the crosstalk model after TNFa and FasL stimulation are the ones associated with Bid and also all reactions associated with Bim (see the supporting information for the model equations).

XIAP has a prominent role as a caspase-3 buffer, and the function of Bel2 family members has turned out to be essential for the model because the sensitizing effect is completely disrupted otherwise (Supporting Fig. 15).

Consequently, it would be of special interest to further analyze the specific function and interplay of pBim and other members of the Bel2 family.

Because many chronic liver diseases in which FasL levels are elevated are associated with chronic