Model-guided identification of a therapeutic strategy to reduce hyperammonemia in liver diseases

(1)

HAL Id: hal-01257127

https://hal.archives-ouvertes.fr/hal-01257127

Submitted on 18 Jan 2016

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Model-guided identification of a therapeutic strategy to

reduce hyperammonemia in liver diseases

Ahmed Ghallab, Geraldine Celliere, Sebastian Henkel, Dominik Driesch,

Stefan Hoehme, Ute Hofmann, Sebastian Zellmer, Patricio Godoy, Agapios

Sachinidis, Meinolf Blaszkewicz, et al.

To cite this version:

Ahmed Ghallab, Geraldine Celliere, Sebastian Henkel, Dominik Driesch, Stefan Hoehme, et al..

Model-guided identification of a therapeutic strategy to reduce hyperammonemia in liver diseases. Journal

of Hepatology, Elsevier, 2015, 64 (4), pp.860-871. �10.1016/j.jhep.2015.11.018�. �hal-01257127�

(2)

Journal: JHEPAT

Article Number: 5901

Please e-mail your responses and any corrections to: E-mail:proofcorrections.esnl@elsevier.sps.co.in

Dear Author,

Please check your proof carefully and mark all corrections at the appropriate place in the proof (e.g., by using on-screen annotation in the PDF file) or compile them in a separate list. Note: if you opt to annotate the file with software other than Adobe Reader then please also highlight the appropriate place in the PDF file. To ensure fast publication of your paper please return your corrections within 48 hours.

For correction or revision of any artwork, please consulthttp://www.elsevier.com/artworkinstructions.

Any queries or remarks that have arisen during the processing of your manuscript are listed below and highlighted by flags in the proof. Click on the ‘Q’ link to go to the location in the proof.

Location in article

Query / Remark:click on the Q link to go

Please insert your reply or correction at the corresponding line in the proof

Q1

Your article is registered as a regular item and is being processed for inclusion in a regular issue of the journal.

If this is NOT correct and your article belongs to a Special Issue/Collection please contact

j.bakker@elsevier.

com

immediately prior to returning your corrections.

Q2

The author names have been tagged as given names and surnames (surnames are highlighted in teal color).

Please confirm if they have been identified correctly.

Q3

The country name has been inserted for the affiliation `6'. Please check, and correct if necessary.

Q4

The affiliations `7' and `11' were identical, the latter has been removed, and hence the affiliation link the author

`Rolf Gebhardt' has been changed, accordingly. Please check, and correct if necessary.

Q5

One or more sponsor names may have been edited to a standard format that enables better searching and

identification of your article. Please check and correct if necessary.

Q6

The country names of the Grant Sponsors are provided below. Please check and correct if necessary. `BMBF'

-`Germany'.

Please check this box if you have no

corrections to make to the PDF file

(3)

Graphical abstract

pp xxx–xxx Model-guided identification of a therapeutic strategy to reduce hyperammonemia in liver diseases

Ahmed Ghallab*_{, Géraldine Cellière, Sebastian G. Henkel, Dominik Driesch,}

Stefan Hoehme, Ute Hofmann, Sebastian Zellmer, Patricio Godoy, Agapios Sachinidis, Meinolf Blaszkewicz, Raymond Reif, Rosemarie Marchan, Lars Kuepfer, Dieter Häussinger, Dirk Drasdo, Rolf Gebhardt, Jan G. Hengstler*

JHEPAT 5901 No. of Pages 1, Model 5G

8 January 2016

(4)

1 2

3

Model-guided identification of a therapeutic strategy to

4

reduce hyperammonemia in liver diseases

5

Ahmed

Ghallab

1,2,⇑

,

Géraldine

Cellière

3,y

,

Sebastian G.

Henkel

4,y

,

Dominik

Driesch

4

,

6

Stefan

Hoehme

5

,

Ute

Hofmann

6

,

Sebastian

Zellmer

7

,

Patricio

Godoy

1

,

Agapios

Sachinidis

8

,

7

Meinolf

Blaszkewicz

1

,

Raymond

Reif

1

,

Rosemarie

Marchan

1

,

Lars

Kuepfer

9

,

Dieter

Häussinger

10

,

8

Dirk

Drasdo

3,5,à

,

Rolf

Gebhardt

7,à

,

Jan G.

Hengstler

1,⇑,à

9 1_{Leibniz Research Centre for Working Environment and Human Factors at the Technical University Dortmund, Dortmund, Germany;}

10 2_{Department of Forensic Medicine and Toxicology, Faculty of Veterinary Medicine, South Valley University, Qena, Egypt;}3_{Institute National de}

11 Recherche en Informatique et en Automatique (INRIA), INRIA Paris-Rocquencourt & Sorbonne Universités UPMC Univ Paris 6, LJLL, France 12 4_{BioControl Jena GmbH, Jena, Germany;}5_{Interdisciplinary Centre for Bioinformatics, University of Leipzig, Leipzig, Germany;}6_{Dr. Margarete}

13 Fischer-Bosch Institute of Clinical Pharmacology and University of Tuebingen, Germany;7_{Institute of Biochemistry, Faculty of Medicine,}

14 University of Leipzig, Leipzig, Germany;8_{Institute of Neurophysiology and Center for Molecular Medicine Cologne (CMMC), University of Cologne,}

15 Robert-Koch-Str. 39, 50931 Cologne, Germany;9_{Computational Systems Biology, Bayer Technology Services GmbH, Leverkusen, Germany;}

16 10Clinic for Gastroenterology, Hepatology and Infectious Diseases, Heinrich-Heine-University, Düsseldorf, Germany

17 18 19

20 Background & Aims: Recently, spatial-temporal/metabolic 21 mathematical models have been established that allow the sim-22 ulation of metabolic processes in tissues. We applied these mod-23 els to decipher ammonia detoxification mechanisms in the liver. 24 Methods: An integrated metabolic-spatial-temporal model was 25 used to generate hypotheses of ammonia metabolism. Predicted 26 mechanisms were validated using time-resolved analyses of 27 nitrogen metabolism, activity analyses, immunostaining and 28 gene expression after induction of liver damage in mice. More-29 over, blood from the portal vein, liver vein and mixed venous 30 blood was analyzed in a time dependent manner.

31 Results: Modeling revealed an underestimation of ammonia con-32 sumption after liver damage when only the currently established 33 mechanisms of ammonia detoxification were simulated. By iter-34 ative cycles of modeling and experiments, the reductive amida-35 tion of alpha-ketoglutarate (

a

-KG) via glutamate dehydrogenase 36 (GDH) was identified as the lacking component. GDH is released 37 from damaged hepatocytes into the blood where it consumes 38 ammonia to generate glutamate, thereby providing systemic

pro-39 tection against hyperammonemia. This mechanism was exploited

40 therapeutically in a mouse model of hyperammonemia by

41 injecting GDH together with optimized doses of cofactors.

42 Intravenous injection of GDH (720 U/kg),

a

-KG (280 mg/kg) and

43 NADPH (180 mg/kg) reduced the elevated blood ammonia

44 concentrations (>200

l

M) to levels close to normal within only

45 15 min.

46 Conclusion: If successfully translated to patients the GDH-based

47 therapy might provide a less aggressive therapeutic alternative

48 for patients with severe hyperammonemia.

49 Ó 2015 European Association for the Study of the Liver. Published

50

52 53

54 Introduction

55 Recent developments have strongly improved our capability to

56 generate information at multiple spatial and temporal scales

57 [1,2]. However, research on disease pathogenesis is hampered

58 by the difficulty to understand the orchestration of individual

59 components. Here, mathematical models help to formalize

rela-60 tions between components, simulate their interplay, and to study

61 processes that are too complex to be understood intuitively[1].

62 This is particularly important when studying the

pathophysiol-63 ogy of metabolic liver diseases, where due to zonation different

64 metabolic processes take place in pericentral and periportal

hep-65 atocytes[3]. To be able to investigate such complex processes we

66 recently established a technique of integrated metabolic

spatial-67 temporal modeling (IM) [4]. These IM integrate conventional

68 metabolic models into spatial-temporal models of the liver lobule

69 [1,4,5]. The present study was motivated by the IM predictions,

70 which proposed that the conventional mechanisms where

71 ammonia is metabolized by urea cycle enzymes in the periportal

72 compartments of the liver lobules and by glutamine synthetase

73 (GS) reaction in the pericentral compartments (Supplementary

Journal of Hepatology 2016 vol. xxxjxxx–xxx

Keywords: Systems biology; Spatio-temporal model; Ammonia; Liver damage; Liver regeneration.

Received 23 March 2015; received in revised form 15 November 2015; accepted 16 November 2015

⇑Corresponding authors. Addresses: Leibniz Research Centre for Working Environment and Human Factors, IfADo – Ardeystr. 67, D-44139 Dortmund, Germany. Tel.: +49 02311084 356; fax: +49 02311084 403 (A. Ghallab) or Leibniz Research Centre for Working Environment and Human Factors, IfADo – Ardeystr. 67, D-44139 Dortmund, Germany. Tel.: +49 02311084 349; fax: +49 02311084 403 (J.G. Hengstler).

E-mail addresses: ghallab@ifado.de (A. Ghallab), hengstler@ifado.de (J.G. Hengstler).

y _{These authors contributed equally to the work.} à_{These authors share co-senior authorship.}

Abbreviations: CCl4, carbon tetrachloride; GDH, glutamate dehydrogenase; AOA,

aminooxy acetate; ALT, alanine transaminase; AST, aspartate transaminase;a-KG, alpha-ketoglutarate; PDAC, 2,6-pyridinedicarboxylic acid.

Q1

Q2

Q3 Q4

(5)

74 Fig. 1) failed to explain the experimental findings[4]. The IM was 75 applied to an experimental scenario, where the entire pericentral 76 and a part of the periportal compartment of the liver lobules were 77 destroyed by a single high dose of the hepatotoxic compound 78 carbon tetrachloride (CCl4). This leads to compromised nitrogen

79 metabolism and hyperammonemia. In the present study, we per-80 formed a series of new experiments accompanied by simulations 81 with novel models to explore the mechanism responsible for the 82 observed discrepancy. Experimentally, the time-resolved analysis 83 of metabolites and metabolic activities after CCl4 intoxication

84 offers good conditions to study ammonia detoxification and pos-85 sible compensatory mechanisms during the damage and regener-86 ation process. Time-resolved analysis of metabolites was 87 performed in the portal vein and heart blood, representing the 88 ‘liver inflow’, and in the liver vein as ‘liver outflow’. These analy-89 ses allowed a precise experimental validation of model predic-90 tions. Finally, iterative cycles of modeling and experimental 91 validation allowed the identification of a so far unrecognized 92 mechanism of ammonia detoxification. Importantly, this mecha-93 nism could be exploited therapeutically to reduce elevated blood 94 ammonia concentrations close to normal levels by intravenous 95 injection of glutamate dehydrogenase (GDH; 720 U/kg) and its 96 cofactors alpha-ketoglutarate (

a

-KG; 280 mg/kg) as well as 97 NADPH (180 mg/kg). This example illustrates how concrete ther-98 apies can be derived by model guided experimental strategies.

99 Materials and methods

100 A detailed description of materials and methods is provided in the Supplemen-101 tarymaterials. Male C57BL/6N10–12weeks old mice were used (Charles River,

102 Sulzfeld, Germany). Acute liver damage was induced by intraperitoneal injection

103 of 1.6 g/kg CCl4, unless other doses are indicated. Blood was taken from mice 104 under anesthesia from the portal and hepatic veins, as well as the right heart

105 chamber, and plasma was separated. Liver tissue samples were collected from

106 defined anatomical positions for histopathology, immunohistochemistry, enzyme

107 activity assays, gene array and q-RT-PCR analyses. The dead cell area was

quan-108 tified in hematoxylin and eosin stained tissue sections using Cell^M software

109 (Olympus, Hamburg, Germany). Whole-genome analysis of gene expression in

110 mouse liver tissue was performed in control as well as after CCl4intoxication 111 with Affymetrix gene arrays. The latter techniques are described fully in the Sup-112 plementary materials and methods. The analysis of ammonia and further

113 metabolites was performed using commercially available kits. Concentrations

114 of amino acids and organic acids in liver tissue were measured in duplicate using

115 GC-MS. GS, GDH and transaminases activity assays were performed

photometri-116 cally as described in theSupplementary materials and methods. NADP+

and

117 NADPH were analyzed by LC-MS. Mouse hepatocytes were isolated by a

two-118 step EGTA/collagenase perfusion technique and either used directly in

suspen-119 sion or cultivated in collagen sandwiches (Supplementarymaterials and meth-120 ods). For the mathematical modeling of ammonia and the related metabolites

121 the integrated metabolic, spatio-temporal model was applied[4,5]. In addition,

122 the IM was replaced by a set of novel models that include further reactions

123 and the blood compartment of the liver (Supplementarymaterials andmethods).

124 Statistical analysis was done with SPSS software as described in the Supplemen-125 tarymaterials.

126 Results

127 An integrated spatial-temporal-metabolic model suggests a so far 128 unrecognized mechanism of ammonia detoxification

129 The detoxification process in healthy, damaged and regenerating 130 livers was simulated using a recently established integrated 131 metabolic IM[4]. To compare the simulated metabolite

concen-132 trations with the in vivo situation, an experiment was performed

133 in which blood was collected from the portal vein (representing

134 85% of the ‘liver inflow’), the heart (representing 15% of the ‘liver

135 inflow’), and the hepatic vein (representing the ‘liver outflow’) in

136 a time-resolved manner after CCl4injection (Fig. 1A;

Supplemen-137 tary Fig.2). The result shows that ammonia is detoxified during

138 its passage through the liver as illustrated by the difference in

139 ammonia concentrations between the portal vein and the hepatic

140 vein in the control mice (Fig. 1B). This detoxification process is

141 compromised after liver damage, particularly on days 1 and 2.

142 Surprisingly, the IM model predicted higher ammonia

concentra-143 tions than those experimentally observed, particularly on day 1

144 (Fig. 1C; see thevideo in the Supplementarydata). Analyses of

145 heart blood demonstrate the contribution of the extrahepatic

146 compartment, which includes brain, muscles, kidneys and blood,

147 to ammonia detoxification between days 1 and 4 after the

induc-148 tion of liver damage. However, this extrahepatic contribution is

149 small compared to detoxification by the liver (Supplementary

150

Figs. 2–8). In addition to the time-resolved study, similar 151 experiments were also performed in a dose dependent manner

152 on day 1 after CCl4administration when the discrepancy between

153 simulated and measured ammonia was maximal. For this

pur-154 pose, doses ranging between 10.9 and 1600 mg/kg CCl4 were

155 tested, resulting in a concentration dependent increase in the

156 dead cell area, with only the highest dose causing damage to

157 the entire CYP2E1 positive pericentral region of the liver lobule

158 (Fig. 1D;Supplementary Fig. 9A,B). Destruction of the GS positive

159 area occurred in with doses ranging between 38.1 and 132.4 mg/

160 kg (Fig. 1D, E; Supplementary Fig. 9C); also CPS1 showed a

161 dose dependent decrease (Supplementary Fig. 9C) leading to

162 compromised ammonia metabolism (Supplementary Fig. 10).

163 Using the IM [4], we also observed a discrepancy between the

164 predicted and measured ammonia in the dose dependent study

165 (Fig. 1F).

166 To find an explanation for this discrepancy, we performed

167 time-resolved gene array analysis of mouse liver tissue after

168 CCl4intoxication (Fig. 2A). Fuzzy clustering identified seven gene

169 clusters which reflected time dependent gene expression

alter-170 ations[6]. Clusters 4 and 6 contained genes whose expression

171 was transiently repressed at early time points after CCl4

intoxica-172 tion (Fig. 2B). Further bioinformatics analyses revealed an over

173 representation of nitrogen/ammonia metabolism KEGG and Gene

174 ontology terms of genes in cluster 4 (Fig. 2C, D).Genes relevant

175 for ammonia metabolism were further studied by qRT-PCR,

176 immunostaining and activity assays. GS is the key enzyme for

177 ammonia detoxification in the pericentral compartment. RNA

178 levels of GS started to decrease as early as 6 h after CCl4injection,

179 it was at its lowest between days 1 and 4, before finally

recover-180 ing to initial levels between days 6 and 30 (Fig. 2E). A similar

181 time-dependent curve was obtained for GS activity although

182 the decrease occurred slightly later than that of RNA with very

183 low levels between days 2 and 4 (Fig. 2E). The pattern and

inten-184 sity of GS immunostaining was found to be comparable to GS

185 activity (Fig. 2F). In addition, ornithine aminotransferase (OAT),

186 an enzyme exclusively localized in GS positive pericentral

hepa-187 tocytes that provides additional glutamate for fixing ammonia

188 [7], decreased to almost undetectable levels with a delayed

189 recovery (Supplementary Fig. 3A). The key enzymes of the

190 periportal compartment, CPS1, ASS1, ASL and arginase1 were

191 similarly analyzed in the same tissue (Supplementary Figs.3B

192 and 4). Extending the IM [4], with time-dependent

Research Article

Please cite this article in press as:GhallabA et al. Model-guided identification of a therapeutic strategy to reduce hyperammonemia in liver diseases. J Hepatol (2016),http://dx.doi.org/10.1016/j.jhep.2015.11.018

2 Journal of Hepatology 2016 vol. xxxjxxx–xxx

JHEPAT 5901 No. of Pages 13

(6)

193 enzyme concentrations (model 1), did not remove the discrep-194 ancy between model predictions and experimental data 195 (SupplementaryFig.11), indicating that our model lacks a rele-196 vant, but so far unrecognized mechanism of ammonia 197 detoxification.

198 Acute liver damage provides systemic protection against ammonia

199 by GDH release

200 Further evidence that an unrecognized mechanism of ammonia

201 detoxification exists arose from metabolic analyses performed

Heart Analyzed parameters • ammonia • urea • glutamine • glutamate • glucose • lactate • pyruvate • alanine • α-ketoglutarate • arginine • other amino acids • ALT, AST

Sites of blood collection

Hepatic vein Portal vein

A

0.0 0 h 1 h 6 h 12 h 1 d 2 d 3 d 4 d 6 d 12 d 30 d 0.5 Ammonia (mM) 0.2 0.3 0.1 0.4 * * * * * _* Portal vein

Time after CCl₄administration

Hepatic vein Heart

B

0.0 0 h 1 h 6 h 12 h 1 d 2 d 3 d 4 d 6 d 12 d 30 d 0.5 Ammonia (mM) 0.2 0.3 0.1 0.4 ** *** *** *** * ** ** meas, in

meas, out sim, out

C

0 h 12 h 24 h 0.2 mM 2 d 4 d 6 d 0.0 mM

D

Control 10.9 mg/kg 38.1 mg/kg 132.4 mg/kg 460 mg/kg 1600 mg/kg Grossly H&E CYP2E1 GS

E

0 0.0 10.9 38.1 132.4 460.0_1600.0 600 1000 * * * GS activity (mU/mg protein) Dose of CCl4 (mg/kg) 200 800 400 0.0 Ammonia (mM) 0.2 0.3 0.1 0.4 ** ** *** meas, in meas, out sim, out 0.0 _10.9 _38.1 132.4 460.0 _1600.0 Dose of CCl4 (mg/kg)

F

JOURNAL OF HEPATOLOGY

(7)

202 using plasma from mice after CCl4injection (Fig. 3A). Most of the

203 analyzed factors in plasma (urea, glutamine, glucose, lactate, 204 pyruvate, alanine, arginine and other amino acids: Supplemen-205 tary Figs. 4–6) were within the expected concentration ranges,

206 _a

-KG, which dramatically decreased between 12 h and day 2 207 (Fig. 3A). This decrease was accompanied by an almost concur-208 rent increase in glutamate levels, which persisted longer than 209 the drop in

a

-KG. One potential explanation is the delayed recov-210 ery of GS, which uses glutamate and ammonia to form glutamine 211 (Fig. 2E, F).The decrease in

a

-KG (and the increase in glutamate) 212 was also accompanied by increased GDH activity in plasma, 213 because GDH is released from damaged hepatocytes (Fig. 3A). 214 The present observations suggest that GDH released from the 215 damaged hepatocytes into the blood catalyzes, at least tran-216 siently, a reaction that consumes ammonia to produce glutamate 217 (Fig. 3D). To test this hypothesis, we collected plasma from mice 218 on day 1 after CCl4injection. Addition of

a

-KG alone was

suffi-219 cient to slightly but significantly decrease blood ammonia con-220 centrations (Fig. 3B). This decrease was enhanced by further 221 adding NADPH and particularly GDH; whereas the GDH inhibitor, 222 PDAC completely antagonized the effect. To test also higher 223 ammonia concentrations typically observed in patients with sev-224 ere pre-coma hyperammonemia, 600

l

M ammonia was added to 225 plasma collected on day 1 after CCl4administration. Under these

226 conditions,

a

-KG also reduced ammonia and increased glutamate 227 concentrations (Fig. 3C;Supplementary Fig.12A). Together, these 228 experiments suggest that a GDH reaction consuming ammonia in 229 blood takes place when GDH is released from acutely damaged 230 livers (Fig. 3D).

231 Validation of the ‘GDH-driven ammonia consumption’ in hepatocytes 232 The experiments described above suggest that high ammonia 233 concentrations in plasma leads to a ‘reverse’ GDH reaction, which 234 consumes rather than produces ammonia. To test whether this 235 ‘GDH-driven ammonia consumption’ occurs not only in plasma 236 but also in cells, we used an in vitro system with primary mouse 237 hepatocytes incubated with ammonia in suspension (Fig. 4). 238 PDAC was used to inhibit GDH (Fig. 4A) in order to determine 239 its influence on ammonia metabolism. In hepatocytes isolated 240 from control mice, unphysiologically high ammonia concentra-241 tions (2 mM) were required until PDAC caused a significant 242 increase of ammonia levels in the suspension buffer (Fig. 4B). 243 However, when hepatocytes from mice 24 h after CCl4

intoxica-244 tion were used, PDAC treatment increased ammonia concentra-245 tions in the suspension buffer, even with 0.5 mM ammonia. 246 Furthermore, in the absence of ammonia, hepatocytes secreted 247 a small but statistically significant amount of ammonia into the 248 buffer. Similarly, glutamate production was reduced by PDAC,

249 an effect that was also stronger in hepatocytes isolated from

250 CCl4-exposed mice (Fig. 4C), which corresponds to the reverse

251 GDH reaction proposed in Fig. 3D (right panel). CCl4 destroys

252 the pericentral hepatocytes (Fig. 1D), which explains the reduced

253 glutamine generation by GS (Fig. 4D) and compromises urea cycle

254 enzymes (SupplementaryFig. 3B,C), which explains the reduced

255 urea production (Fig. 4E). Similar experiments were also

per-256 formed with cultivated (instead of suspended) hepatocytes from

257 untreated mice. The results demonstrate that inhibition of GDH

258 at high ammonia concentrations increases ammonia-induced

259 cytotoxicity (Supplementary Fig.12B). These results show that

260 the catalytic direction of GDH reverses a clearly becomes

ammo-261 nia consuming also in hepatocytes in order to compensate the

262 compromised metabolism by urea cycle enzymes and GS after

263 intoxication.

264 Further evidence emerges from simulations with a set of

265 novel models 1–4 (SupplementaryFig.11). If a reversible GDH

266 reaction was integrated into the hepatocyte compartment

267 (Fig. 5A; Supplementary Fig. 11), the discrepancy between

268 in vivo measured and simulated ammonia concentrations

269 (Fig. 1C) completely disappeared (Fig. 5B). The quantitative

270 agreement was obtained even without considering the blood

271 compartment of the liver, suggesting that after CCl4-induced

272 damage, the ammonia consumption catalyzed by GDH in

273 the hepatocytes represents the missing ammonia sink predicted

274 by[4].

275 Therapy of hyperammonemia based on the reverse GDH reaction

276 The above described ammonia consumption catalyzed by the

277 GDH reaction (Figs. 3B–D and 4) and the aforementioned

278 decrease in plasma

a

-KG levels (Fig. 3A) prompted us to test

279 whether supplementation of

a

-KG in mice helps to detoxify

280 ammonia. Therefore, mice received a hepatotoxic dose of CCl4

281 (1.6 g/kg) and 24 h later

a

-KG (280 mg/kg) was injected into

282 the tail vein. Blood was collected immediately before as well as

283 15, 30 and 60 min after injection of

a

-KG. A decrease in plasma

284 ammonia concentrations by 31, 40 and 43% was observed 15,

285 30 and 60 min after

a

-KG injection, respectively (Fig. 6A).

Gluta-286 mate increased after 15 min and decreased again after longer

287 periods probably due to the consumption by further metabolism.

288 a

-KG transiently increased in plasma after injection and then

289 rapidly decreased. Analysis of GDH activity demonstrated that

290 the experiment was performed under conditions of high plasma

291 activity. In control mice, injection of

a

-KG did not alter blood

292 concentrations of ammonia or glutamate (Fig. 6B). In addition,

293 plasma

a

-KG levels were lower in CCl4-treated mice compared

294 to the control mice, suggesting increased consumption in mice

295 with damaged livers.

Fig. 1. Evidence for a so far unrecognized mechanism of ammonia detoxification. (A) Experimental design. (B) Ammonia concentrations in the portal vein, hepatic vein and heart.⁄p <0.05 compared to the corresponding controls (0 h). (C) Integrated metabolic spatio-temporal model using the technique described by[4](video in the Supplementary data). Predicted ammonia concentrations in the liver outflow are higher compared to the experimental data.⁄⁄⁄p <0.001,⁄⁄p <0.01 and⁄p <0.05 compared to the measured ammonia output. (D) Dose dependent experiment (10.9 to 1600 mg/kg CCl424 h after administration) showing macroscopic alterations with a spotted

pattern at 132.4 mg/kg and higher doses, corresponding to the central necrotic lesion in hematoxylin/eosin staining, scale bars: 100lm. Destruction of the pericentral CYP2E1 positive region which begins at 132.4 mg/kg with central necrosis still surrounded by CYP2E1 positive surviving hepatocytes; the entire CYP2E1 positive region was destroyed at the highest dose of 1600 mg/kg. The GS positive region was destroyed only at 132.4 mg/kg and higher doses, which corresponds to the decrease in GS activity (E), scale bars: 200lm.⁄p <0.05 when compared to the control group (0). (F) Comparison of analyzed and simulated ammonia concentrations in the liver vein for the experiment in (D); meas, in: analyzed concentrations in the portal vein (representing 85% of the liver inflow) and heart blood (representing 15% of the liver inflow); meas. out: analyzed concentrations in the liver vein; sim. out: simulated concentrations in the liver vein. Data are mean values and SD of three mice per time point and dose of CCl4.⁄⁄⁄p <0.001 and⁄⁄p <0.01 compared to the measured ammonia output.

Research Article

(8)

0 1 2 3 4 6 30 CCl4 (1.6 g/kg)

A

B

0 10 15 0.0 -1.0 -0.5 0.5 1.0 5 Cluster 4 310 genes Fold change

Time after CCl₄ administration (days) KEGGnitrogenmetabo mils Time after CCl₄administratio n

l o b m y

S Name Healthy _liver Hour 2 Hour 8 Day 1 Day 2 Day 4 Day 6 Day 8 Day 16

Car1 Carbonicanhydrase 1 1 -1.12 -1.78 -2.42 -1.85 -1.11 1.1 1.15 1.02 Car3 Carbonicanhydrase 3 1 -2.94 -4.51 -32.55 -4.47 -1.33 -1.74 -1.71 -1.41 Car5a Carbonic anhydrase 5a, _m_i_t_o_c_h_o_n_d_r_i_l_a 1 -1.64 -4 -4.44 -1.96 1.1 -1.17 1.02 -1.1 Cth _{(cystathionine}Cystathionase _-_l_y_a_s_e₎ 1 -1.28 -1.95 -2.59 -2.03 1.35 -1.08 -1.04 -1.09 Gls2 ₍_il_vG_el_ru_,t_mam_i_t_oin_c_has_oe_n 2_d_r₎_i_a_l 1 -1.65 -2.44 -3.71 -2.13 -1.06 -1.26 -1.01 -1.14 Glul Glutamate-ammonia ligase ₍_g_l_u_t_a_m_i_n_e_s_y_n_t_h_e_t_a₎_s_e 1 1.13 -1.84 -5.82 -4.85 -2.36 -1.02 -1.27 -1.08 Hal Histidineammonialyas e 1 -1.71 -3.09 -3.86 -2.11 1.05 -1.37 1.14 1

C

D

_G_O_u_r_e_a_c_y_c_l_e_/_u_r_e_a_m_e_t_a_b_._p_r_o_c_e_s_s _{Time after CCl} 4administratio n l o b m y S Name Healthy

liver Hour 2 Hour 8 Day 1 Day 2 Day 4 Day 6 Day 8 Day 16 Aldh6a1 Aldehyde dehydrogenase_f_a_m_li_y₆_,_s_u_b_f_a_m_li_y₁_A 1 -1.15 -1.63 -2.33 -2.09 -1.24 1.02 1.02 -1.03

Dpyd Dihydropyrimidine _d_e_h_y_d_r_o_g_e_n_a_s_e 1 -1.28 -1.71 -2.63 -2.15 -1.27 -1.08 1.07 -1.02 Dpys Dihydropyrimidina es 1 -1.26 -2.91 -3.35 -1.87 -1.08 -1.01 -1.05 -1.12 Asl Argininosuccinatelyas e 1 1.06 1.5 -2.12 -1.74 -1.13 -1.13 -1.05 -1.09 Cebpa CCAAT/enhancer binding _p_r_o_t_e_i_n₍_C_/_E_B_P₎_,_a_l_p_h_a 1 -1.25 -1.35 -2.07 -1.4 1.06 1.14 1.05 -1.07 Otc _{transcarbamylase}Ornithine 1 -1.22 -1.89 -3.26 -2.14 1.06 -1.08 1.01 -1.01

E

0.0 0.6 1.2 * * * * * * * * 0.9 0.3 0 h 1 h 6 h 12 h 1 d 2 d 3 d 4 d 6 d 12 d 30 d GS RNA (2 -ΔΔct )

Time after CCl4 administration 0 150 250 * _* _* * * 200 50 100 0 h 1 h 6 h 12 h 1 d 2 d 3 d 4 d 6 d 12 d 30 d GS activity (U/liver)

F

12 h 4 d 30 d Control 1 d 6 d Time (days)

Fig. 2. Spatio-temporal alterations of ammonia metabolizing enzymes after CCl4intoxication. (A) Experimental design. (B) Time dependent changes of gene expression

in fuzzy cluster 4 from[6]. The dots correspond to the average of the mean scaled values for all 310 genes, between their respective maximal and minimal expression levels at each time point, using healthy liver (time 0) as reference. Error bars indicate standard error. (C) Changes in expression of genes associated to the KEGG terms ammonia/ nitrogen metabolism (Gene Ontology [GO] ID 910) as revealed by KEGG pathways enrichment analysis in fuzzy cluster 4 (p = 2.36 107_{). (D) Changes in the expression of}

genes associated to the GO terms ‘urea cycle/urea metabolic process’ (Gene Ontology ID 0000050 and 0019627 respectively) as revealed by GO enrichment analysis in fuzzy cluster 4 (p = 3.83 104_{). In C and D, the values indicate fold of expression over healthy liver at each time point after CCl}

4administration, and correspond to the average of

5 independent biological replicates. Time course of GS RNA levels, GS activity (E) and immunostaining (F), Scale bars: 200lm.⁄p <0.05 when compared to the control group (0 h).

JOURNAL OF HEPATOLOGY

(9)

0 0 h 1 h 6 h 12 h 1 d 2 d 3 d 4 d 6 d 12 d 300 GDH activity (U/I) 200 100 Portal vein Hepatic vein Heart * * * _* 0 0 h 1 h 6 h 12 h 1 d 2 d 3 d 4 d 6 d 12 d 45 α-ketoglutarate (μM) 30 15 * * * * * * * * * _* _* * 0 0 h 1 h 6 h 12 h 1 d 2 d 3 d 4 d 6 d 12 d 200 Glutamate (μM) 150 100 50

* *

A

B

0 150 200 * * * Ammonia (μM) 50 100 0 600 _* * * * Glutamate (μM) 200 400 0 6000 *** GDH activity (U/I) 2000 4000 α-KG - + + + + + AOA - - + + + + NADPH - - - + + + GDH - - - - + + PDAC - - - + 0 * Ammonia (μM) 1000 400 200 600 800 0 ** Glutamate (μM) 500 200 100 300 400 0 GDH activity (U/I) 300 200 100 *** Control NH4Cl NH₄Cl + α-KG NH4Cl + α-KG + PDAC

C

D

Normal situation Liver Blood Brain (Low; <50 μM) Pericentral Periportal Glnase Gln Gln GS GDH GDH Urea Urea cycle Glu Glu NH4 + NH₄+ NH4 + NH₄+ α-KG_in α-KG_out α-KG α-KG α-KG + No toxicity

After liver damage: depletion of α-KG

(Hyperammonemia) (depletion) Pericentral Periportal Glnase (depletion) (compromised) Gln Gln GS GDH GDH GDH GDH Urea Urea cycle Glu Glu NH₄+ NH₄+ NH4 + NH₄+ α-KG_in α-KG_out α-KG α-KG + Risk of hepatic encephalopathy

X

GDH Glu α-KG *

Fig. 3. Detoxification of ammonia by a reverse GDH reaction. (A) After induction of liver damage by CCl4, plasma activity of GDH transiently increases. This is

accompanied by a decrease in alpha-ketoglutarate (a-KG) and an increase in glutamate.⁄p <0.05 when compared to the corresponding control (0 h). Similar results were observed in liver tissue (Supplementary Tables 1 and 2). (B) Validation of the reverse GDH reaction using an inhibitor of GDH (PDAC). Plasma of mice 24 h after CCl4

injection was analyzed.a-KG was added alone or in combination with AOA, NADPH, GDH and PDAC.⁄⁄⁄_{p <0.001,}⁄⁄_{p <0.01 and}⁄_{p <0.05 when compared to controls (0). (C) A}

similar experimental design was chosen as in B. However, 600lM ammonia was added to also test the reaction at a higher but still clinically relevant concentration.a-KG decreases ammonia and increases glutamate concentrations, which can be blocked by the GDH inhibitor, PDAC.⁄⁄⁄p <0.001,⁄⁄p <0.01 and⁄p <0.05 when compared to the NH4Cl group (0 h). Data are mean values and SD of 3 biological replicas. (D) Concept of the reverse glutamate dehydrogenase (GDH) reaction. In normal periportal

hepatocytes, GDH generates ammonia, which is detoxified by the urea cycle. In pericentral hepatocytes, GDH generates glutamate which is required as a substrate for the GS reaction to form glutamine (Gln). Biosynthesis ofa-KG takes place in the periportal hepatocytes;a-KG is then partially exported and taken up again by the pericentral hepatocytes, where it is needed for GS[25,26]. After induction of liver damage, the expression of urea cycle enzymes decreased and the pericentral region with GS is completely destroyed. This leads to increased blood ammonia concentrations. However, also GDH is released from damaged hepatocytes and catalyzes a reaction in blood consuming ammonia anda-KG to generate glutamate (Glu). This reaction can go untila-KG in blood is consumed. In this situationa-KG and NADPH should be therapeutically substituted.

Research Article

(10)

296 In the aforementioned experiment, the molar amount of glu-297 tamate produced in the damaged liver after

a

-KG injection was 298 higher than ammonia consumption (Fig. 6A). Therefore, the 299 results cannot only be explained by the reverse GDH reaction, 300 but may be due to the consumption of

a

-KG by transaminases 301 that contribute to the generation of glutamate. Indeed, tail vein 302 injection of the transaminases inhibitor AOA prior to

a

-KG injec-303 tion reduced the production of glutamate (Fig. 6C) and improved 304 ammonia detoxification. The efficiency of transaminases inhibi-305 tion by AOA in vivo has been confirmed in preliminary experi-306 ments (Supplementary Figs.13 and14).

307 The reverse GDH reaction requires NADPH as a cofactor; how-308 ever, NADPH concentrations are very low in blood. To determine 309 how NADPH levels are altered in our model of liver damage, both 310 NADPH and its oxidized form NADP+_{were analyzed. Blood}

con-311 centrations of both NADPH and NADP+_{increased after induction}

312 of liver damage by CCl4(SupplementaryFig.15A). In addition, an

313 enhanced NADP+_{/NADPH ratio was observed in both blood and}

314 liver tissue (Supplementary Fig.15B). This increase in NADP+_/

315 NADPH ratio fits to a switch in the GDH reaction from NADPH

316 generation to NADPH consumption. Despite the increase in

317 NADPH after induction of liver damage, the concentrations are

318 still relatively low. Therefore, to study the influence of NADPH,

319 plasma from mice collected 24 h after CCl4injection was

incu-320 bated with varying concentrations of NADPH in the presence of

321 NH4Cl (1 mM),

a

-KG (3 mM), AOA (1 mM) and GDH (12,000 U/

322 l) for one hour. A concentration dependent decrease in plasma

323 ammonia and an increase in glutamate were observed with

324 increasing concentrations of NADPH (Fig. 7A). A similar trend

325 for ammonia and glutamate was observed with increasing

con-326 centrations of

a

-KG and GDH (Fig. 7B, C).Moreover, addition of

327 AOA reduced both ammonia and glutamate concentrations (

Sup-328

plementaryFig.16). To understand how the orientation of the 329 GDH reaction is controlled by ammonia and glutamate

concen-330 trations, titration experiments were performed, which indicated

331 that GDH significantly consumes ammonia beginning at

concen-332 trations of 150

l

M and higher (Fig. 7D). In contrast,

unphysiolog-333 ically high concentrations of more than 10 mM glutamate were

334 required to block the reaction (Fig. 7E).

335 Based on these in vitro optimized concentrations, we designed

336 an in vivo study to treat hyperammonemia in mice. After the

337 induction of liver damage by CCl4, transaminases activities were

338 inhibited by AOA (13 mg/kg; tail vein injection; 24 h after CCl4

339 administration). Thirty minutes later a cocktail of

a

-KG

340 (280 mg/kg), GDH (720 U/kg) and NADPH (180 mg/kg) was

intra-341 venously injected. A dose of 280 mg/kg

a

-KG was chosen because

A

* ** * 0 15 25 *** ** **

GDH activity (U/g protein)

5 20 10 Control CCl4 (1.6 g/kg) -PDAC +PDAC -PDAC +PDAC ** ºº ** º * 0 1500 2500 * Ammonia (μM) 500 2000 1000

B

C

0 0.0 60 ** * Glutamate (μM) 20 40 0.5 2.0 NH4Cl (mM) * * * º º º 0.0 0.5 2.0 NH4Cl (mM)

D

0 80 ** Glutamate (μM) 20 40 60 Control CCl₄

E

0 250 * Urea (μM) 50 100 150 200 Control CCl₄

Fig. 4. Ammonia consumption by GDH in primary mouse hepatocytes. Hepatocytes were isolated from CCl4(1.6 g/kg) intoxicated (day 1) and untreated

mice and suspended at a concentration of 2 million hepatocytes/ml for 1 h with different concentrations of ammonia. (A) Inhibition of GDH activity by PDAC. (B) Compromised ammonia detoxification after GDH inhibition. (C) Reduced gluta-mate production by GDH inhibition.⁄⁄⁄p <0.001,⁄⁄p <0.01 and⁄p <0.05 compared to – PDAC.°°p <0.01 and°p <0.05 compared to hepatocytes from untreated mice. (D & E) compromised urea and glutamine production by hepatocytes of CCl4

intoxicated mice.⁄⁄p <0.01 and⁄p <0.05 compared to hepatocytes from untreated mice. Data are mean values and SD of three independent experiments.

A

Blood far periportal Cellular far

periportal Cellularmid-periportal Cellular pericentral GLNase GDH CPS Gln Gln Urea Glu α-KG NH4

Blood mid-periportal Blood pericentral GDH GDH CPS GS Urea Glu Glu α-KG α-KG NH4 NH₄ 0 0 h 1 h 6 h 12 h 1 d 2 d 3 d 4 d 6 d 12 d 500 Ammonia (μM) 200 300 100 400 meas, in

Time after CCl4 administration

meas, out sim, out

B

Fig. 5. Integration of the GDH reaction into the metabolic model. (A) Scheme of the metabolic reactions and zones of the extended model including GDH in the blood of the liver and hepatocytes. (B) The model extension leads to a better fit between simulated and experimental data.

JOURNAL OF HEPATOLOGY

(11)

342 it transiently normalized

a

-KG levels in mice 24 h after CCl4.

343 720 U/kg GDH was used because it resulted in plasma levels of 344 approximately 6000 U/l 15 min after injection (Supplementary

345 Fig.17), an activity level shown to allow maximal ammonia con-346 sumption in plasma in vitro (Fig. 7C). The dose of 180 mg/kg 347 NADPH was also considered as adequate in a pharmacokinetic 348 experiment (Supplementary Fig. 18) as it transiently increased 349 plasma NADPH to approximately 1.6 mM 2 min after injection. 350 Injection of the

a

-KG/GDH/NADPH cocktail (KGN cocktail) 351 reduced ammonia concentrations from 213 to 74

l

M within 352 15 min after administration (Fig. 8). Simultaneously, glutamate 353 levels increased from 131 to 369

l

M. Analysis of

a

-KG and 354 GDH activity in the plasma showed that substitution was suc-355 cessful 15 min after the injection of the KGN cocktail. Moreover, 356 the activities of aspartate and alanine aminotransferase were suc-357 cessfully inhibited by AOA. The mice were observed for three 358 weeks after the experiment and did not show any complications.

359 Discussion

360 Guided by simulations with an IM predicting a missing ammonia 361 sink after severe CCl4-induced liver damage, we identified the

362 GDH reaction as fundamental for ammonia consumption, which

363 can be used therapeutically by the administration of a cocktail

364 of GDH and cofactors.

365 Therapy for hyperammonemia remains challenging [8–10].

366 Hemodialysis is the most efficient treatment for reducing

ele-367 vated blood ammonia concentrations [11,12]. For milder forms

368 of hyperammonemia, pharmacologic management is possible

369 [13]. Efficient strategies for patients with urea cycle defects

370 include infusion of phenylacetate or benzoate. Phenylacetate

371 combines with glutamine to form a product which can be

372 excreted by the kidneys[9,13,14]. Conversely, benzoate combines

373 with glycine to form hippurate, which is also excreted in urine

374 [13,15]. Both compounds reduce the total body nitrogen content;

375 however, this therapy has also failed in a fraction of patients with

376 hyperammonemic crisis who became refractory most probably

377 due to the accumulation of nitrogen waste [9]. This led to the

378 concept that only blood ammonia concentrations below 500

l

M

379 should be treated pharmacologically; whereas, more severe

380 hyperammonemia requires aggressive interventions with renal

381 replacement therapies, such as hemodialysis[12,16]. In such

sit-382 uations with either severe or refractory hyperammonemia the

383 therapeutic strategy developed in the present study may be an

384 alternative to hemodialysis.

385 The current experiments demonstrate that infusion of a

KGN-386 cocktail reduces ammonia close to normal levels within minutes.

A

CCl

4

treated mice

Time after α-KG injection (min) 0 0 15 30 60 150 450 ** * Ammonia (μM) 300 0 0 15 30 60 150 450 * *** Glutamate (μM) 300 0 0 15 30 60 20 80 α-Ketoglutarate (μM) 40 60 0 0 15 30 60 100 400 GDH activity (U/I) 200 300

B

Control mice 0 0 15 30 60 30 90 Ammonia (μM) 600 0 0 15 30 60 20 80 Glutamate (μM) 40 60 0 0 15 30 60 300 1200 _*** * α-Ketoglutarate (μM) 600 900 0 0 15 30 60 20 40 GDH activity (U/L)

Time after α-KG infusion (min)

C

CCl 4 and AOA treated mice

Time after α-KG injection (min) 0 300 Ammonia (μM) 100 200 -30 0 15 0 800 * Glutamate ( μ M) 200 400 600 -30 0 15 ** 0 1000 α-Ketoglutarate (μ M) 200 400 600 800 -30 0 15 0 Glutamate (μ M) 100 200 300 -AOA +AOA -30 0 15

Fig. 6. Reduction of blood ammonia concentrations bya-KG. (A) Tail vein injection of 280 mg/kga-KG into mice 24 h after induction of liver damage by CCl4(1.6 g/kg).

(B) Control experiment witha-KG (280 mg/kg) injected into the tail vein of untreated mice.⁄⁄⁄p <0.001,⁄⁄p <0.01 and⁄p <0.05 when compared to the control group (0). (C) Influence of the transaminase inhibitor, AOA (13 mg/kg; tail vein injection) on ammonia detoxification bya-KG. ⁄⁄p <0.01 and⁄p <0.05 when compared to the corresponding control. Data are mean values and SD of three mice treated at different experimental days with individually prepareda-KG.

Research Article

(12)

387 Under these conditions, a GDH-catalyzed reaction takes place in 388 blood where ammonia and

a

-KG are consumed to form gluta-389 mate in an NADPH-dependent reaction. GDH was also previously 390 reported to switch its catalytic orientation under physiological 391 conditions. In the periportal compartment of the liver lobule, 392 GDH generates ammonia (Fig. 3D), which fuels the urea cycle. 393 In the pericentral compartment, GDH is known to consume 394 ammonia to generate glutamate for the GS reaction [17–19].

395 The present study shows by use of a GDH inhibitor that GDH

396 released from damaged liver tissue may catalyze an ammonia

397 consuming reaction under conditions of hyperammonemia. By

398 releasing GDH from damaged hepatocytes into the blood, the

399 damaged liver provides a mechanism that reduces blood

ammo-400 nia levels. However, this protective mechanism is limited by the

401 availability of the GDH substrate,

a

-KG. The present study shows

402 that

a

-KG strongly decreases upon induction of acute liver

dam-A

0 300 1500 ** *** *** *** *** Ammonia (μM) 900 600 1200 0 - + + + 300 1500 *** *** *** _*** *** Glutamate (μM) + + + + 900 600 1200 NH4 Cl (1 mM) - - + + + + + + Cofactors - - -125 250 _{500 1000 2000} NADPH ( μM)

B

0 500 1500 *** *** *** *** Ammonia (μM) 1000 0 - + + + 500 * ** *** Glutamate (μM) + + + 1500 1000 2000 - - + + + + + - - -375 _{750 1500 3000} NH4 Cl (1 mM) Cofactors α-KG (μM)

C

0 300 1500 * ** *** *** *** *** Ammonia (μM) 900 600 1200 0 - + + + 300 *** ** *** *** *** _*** *** Glutamate (μM) + + + 900 1200 600 1500 - - + + + + + - - -188 375 _{750 1500} + + 3000 + + 6000 + + 12,000 NH4 Cl (1 mM) Cofactors GDH (U/L)

D

0 750 Glutamate ( μM) 150 450 300 600 NH₄Cl (μM) 37.5 75 150 300 600 *** *** -Cocktail +Cocktail 0 750 Ammonia (μM) 150 450 300 600 -Cocktail +Cocktail * _** *** 37.5 75 150 300 600

E

0 200 400 800 *** ** ** *** Ammonia (μM) 600 0 10 20 40 Glutamate (mM) 30 + + + + + + - + + + + + - -3.16 10.00 31.60 NH₄Cl Cocktail Glutamate (mM) 1.00

Fig. 7. Optimization of cofactor concentrations for the GDH reaction and identification of maximal GDH activity. (A) NADPH: mouse plasma collected 24 h after CCl4

injection was incubated with varying concentrations of NADPH in the presence of NH4Cl (1 mM) and other cofactors; aminooxy acetate (AOA) (1 mM),a-KG (3 mM) and

GDH (12,000 U/L) for one hour. (B) Alpha-ketoglutarate (a-KG): plasma was collected from mice on day one after CCl4administration and incubated with varying

concentrations ofa-KG in the presence of NH4Cl (1 mM) and other cofactors; AOA (1 mM), NADPH (1 mM) and GDH (12,000 U/L). (C) GDH: plasma from mice collected 24 h

after CCl4injection was incubated with varying concentrations of GDH in the presence of NH4Cl (1 mM) and other cofactors; AOA (1 mM),a-KG (3 mM), and NADPH (1 mM)

for one hour.⁄⁄⁄p <0.001,⁄⁄p <0.01 and⁄p <0.05 compared to the situation where the plasma was incubated with only NH4Cl. (D) Ammonia detoxification by the cocktail.

Different ammonia concentrations were added to plasma of untreated mice and a cocktail ofa-KG (3 mM), NADPH (0.5 mM), GDH (6000 U/L) and AOA (1 mM) was added to analyze ammonia and glutamate one hour later.⁄⁄⁄p <0.001,⁄⁄p <0.01 and⁄p <0.05 compared – cocktail. (E) Ammonia (600lM) detoxification by the cocktail was only blocked by unphysiologically high glutamate concentrations.⁄⁄⁄p <0.001 compared to NH4Cl alone. Data are mean values and SD of three biological replicas.

JOURNAL OF HEPATOLOGY

(13)

403 age, because it is consumed by the reverse GDH reaction. This 404 prompted us to supplement

a

-KG, increase blood concentrations 405 of NADPH, and infuse GDH. Indeed, the combined injection of

a

-406 KG, GDH and NADPH efficiently reduced blood ammonia concen-407 trations. A GDH bolus was injected to result in plasma peak con-408 centrations between 5000 and 6000 U/L. This is in the same order 409 of magnitude as observed in patients after acetaminophen intox-410 ication[20]. Therefore, in patients with acute liver intoxication 411 with high blood GDH, therapy with

a

-KG and NADPH might be 412 sufficient. However, it should be considered that the GDH reac-413 tion in blood described here (Fig. 3D) does not explain all exper-414 imental observations:

a

-KG decreases significantly 12 h after CCl4

415 administration when there is no significant increase in blood 416 GDH (Fig. 3A). This discrepancy may be explained by the intracel-417 lular change of the catalytic direction of GDH in the periportal 418 hepatocytes, which may precede the GDH release into the blood. 419 However, this was not further analyzed in the present study as

420 we choose to focus on the therapeutically more relevant GDH

421 reaction taking place in blood.

422

a

-KG was previously tested for the treatment of

hyperam-423 monemia between 1964 and 1978 [21,22], but this strategy

424 was abandoned because it was not sufficiently efficient for

clini-425 cal application. The reverse GDH reaction in the blood and its

426 requirement for NADPH as a cofactor was not yet known when

427 the early therapeutic studies with

a

-KG were performed. In

addi-428 tion, injection of

a

-KG alone in the present study resulted in only

429 a relatively weak reduction of hyperammonemia. This reduction

430 may be possible because in addition to GDH, NADPH is also

431 released from deteriorating hepatocytes after CCl4injection.

432 The concept of treating hyperammonemia by

a

-KG/GDH/

433 NADPH infusion originates from simulations using an integrated

434 metabolic spatio-temporal model [4]. This model is based on

435 well-understood pathways of ammonia detoxification, such as

436 urea cycle enzymes and the GS reaction [23,24]. It predicted

437 higher ammonia concentrations compared to the measured data.

438 Therefore, we analyzed liver tissue during the damage induction

439 and regeneration processes, but the results could also not explain

440 the discrepancy. Time-resolved gene array experiments following

441 CCl4 injection led to the observation that a general decrease in

442 metabolizing enzymes occurs including enzymes involved in

443 ammonia metabolism. All enzymes of the urea cycle were

tran-444 scriptionally downregulated by at least 60%. Factors identified

445 by the gene array analysis were further analyzed by activity

446 assays and immunostaining. Key observations were: (a) the GS

447 positive region, which is initially completely destroyed by CCl4,

448 shows a delayed recovery and does not return to normal levels

449 before day 12; and (b) CPS1, the rate limiting enzyme in the urea

450 cycle normally expressed in the periportal region, is

downregu-451 lated during the destruction process (days1–3),but its

expres-452 sion then extends throughout the entire liver lobule during

453 days4–6.The other urea cycle enzymes showed a similar time

454 course as CPS1 with the exception of arginase1, which decreased

455 only slightly during the destruction and regeneration process.

456 Glutaminase showed a similar time course and pattern as

457 CPS1. Nevertheless, none of these alterations could explain the

458 observed discrepancy. However, the refined models that take into

459 account the reversible GDH reaction show an excellent

agree-460 ment with the experimental data suggesting that consumption

461 by the GDH reaction represents the previously predicted

ammo-462 nia sink, hence providing an example for model guided

463 experimentation.

464 In conclusion, a novel form of therapy has been identified that

465 allows the rapid correction of hyperammonemia by the infusion

466 of

a

-KG, GDH and NADPH. This pharmacotherapy may prove

rel-467 evant as an emergency therapy for episodes of hyperammonemia

468 in urea cycle disease or liver cirrhosis.

469 Financial support

470 This study was supported by the BMBF funded projects Virtual

471 Liver(FKZ0315739),Lebersimulator and FP 7EUNOTOX.

472 Conflict of interest

473 The authors who have taken part in this study declared that they

474 do nothave anythingto disclose regarding funding or conflict of

475 interest with respect to this manuscript.

0 50 100 520 Ammonia (μM) 150 200 *** 0 200 400 600 Glutamate (μM) *** - - + - + + KGN-cocktail AOA 0 300 600 900 1200 α-Ketoglutarate ( μ M) Time after KGN injection (min) * -30 0 15 0 2000 4000 8000 AST (U/L) 6000 *** *** 0 4000 8000 16000 12000 AL T (U/L) * * - - + - + + KGN-cocktail AOA 0 2000 4000 6000 GDH activity (U/L) Time after KGN injection (min) *** -30 0 15

A

Fig. 8. Treatment of hyperammonemia by injection of a cocktail of GDH and optimized cofactor doses. A cocktail of GDH (720 U/kg),a-KG (280 mg/kg) and NADPH (180 mg/kg) (KGN) was injected into mice 24 h after induction of liver damage using CCl4(1.6 g/kg). Thirty minutes prior to treatment with the cocktail,

mice received a single dose of 13 mg/kg AOA to block transaminases. Injection of the KGN cocktail reduced ammonia and increased glutamate concentrations in the blood of mice.a-KG and GDH activity increased while aspartate (AST) and alanine aminotransferase (ALT) activities decreased. Data are mean values and SD of four mice treated at different experimental days with individually prepared therapeutic cocktails.⁄⁄⁄_{p <0.001,}⁄⁄_{p <0.01 and}⁄_{p <0.05 when compared to the}

control group (30).

Q5 Q6

Research Article

(14)

476 Authors’ contributions 477

478 Ahmed Ghallab: study concept and design; acquisition of 479 data; analysis and interpretation of data; drafting of the 480 manuscript; statistical analysis; critical revision of the 481 manuscript.

482 Sebastian G. Henkel: mathematical modeling (integrated 483 model of Schliess etal.(2014), model 0 and extrahepatic mass 484 balance); analysis and interpretation of data; drafting of the 485 manuscript

486 Géraldine Cellière: mathematical modeling (novel ammonia 487 detoxification models, extension of model 0); analysis and 488 interpretation of data; drafting of the manuscript, statistical 489 analysis.

490 Dominik Driesch: mathematical modeling (integrated model 491 of Schliess etal.(2014), model 0 and extrahepatic mass bal-492 ance); analysis and interpretation of data; drafting of the 493 manuscript.

494 Stefan Hoehme: mathematical modeling (integrated model of 495 Schliess etal.(2014), model 0); analysis and interpretation of 496 data; drafting of the manuscript.

497 Ute Hofmann: acquisition of data; analysis and interpretation 498 of data; technical support; drafting of the manuscript. 499 Sebastian Zellmer: study concept and design; acquisition of 500 data; analysis and interpretation of data; drafting of the 501 manuscript; critical revision of the manuscript.

502 Patricio Godoy: acquisition of data; analysis and interpreta-503 tion of data; drafting of the manuscript.

504 Agapios Sachinidis: acquisition of data; analysis and interpre-505 tation of data; drafting of the manuscript.

506 Meinolf Blaszkewicz: acquisition of data; analysis and inter-507 pretation of data; technical support; drafting of the 508 manuscript.

509 Raymond Reif: analysis and interpretation of data; drafting of 510 the manuscript.

511 Rosemarie Marchan: critical revision of the manuscript. 512 Lars Kuepfer: mathematical modeling; drafting of the 513 manuscript.

514 Dieter Häussinger: study concept and design; critical revision 515 of the manuscript.

516 Dirk Drasdo: study concept and design; mathematical model-517 ing (integrated model of Schliess etal.(2014), model 0 and 518 novel models); analysis and interpretation of data; drafting 519 of the manuscript, statistical analysis, critical revision of the 520 manuscript

521 Rolf Gebhardt: study concept and design; acquisition of data; 522 analysis and interpretation of data; drafting of the manuscript, 523 critical revision of the manuscript.

524 Jan G. Hengstler: study concept and design; acquisition of 525 data; analysis and interpretation of data; drafting of the 526 manuscript, statistical analysis, critical revision of the manu-527 script; study supervision.

528

529 Acknowledgments

530 We thank Mrs. Gabi Baumhoer, Mrs. Georgia Günther and Mrs. 531 Brigitte Begher-Tibbe – Leibniz Research Centre for Working 532 Environment and Human Factors at the Technical University 533 Dortmund, Dortmund, Germany; and Ms. Margit Henry and Ms. 534 Tamara Rotshteyn – Institute of Neurophysiology and Center

535 for Molecular Medicine Cologne (CMMC), University of Cologne,

536 Cologne,Germany –for competent technical assistance.

537

Supplementary data

538 Supplementary data associated with this article can be found, in

539 the online version, at http://dx.doi.org/10.1016/j.jhep.2015.11.

540 018.

541 References

542 Author names in bold designate shared co-first authorship

543

[1] Drasdo D, Hoehme S, Hengstler JG. How predictive quantitative modelling of

544

tissue organisation can inform liver disease pathogenesis. J Hepatol

545

2014;61:951–956.

546

[2] Zellmer S, Schmidt-Heck W, Godoy P, Weng H, Meyer C, Lehmann T, et al.

547

Transcription factors ETF, E2F, and SP-1 are involved in

cytokine-548

independent proliferation of murine hepatocytes. Hepatology 2010;52:

549

2127–2136.

550

[3] Godoy P, Hewitt NJ, Albrecht U, Andersen ME, Ansari N, Bhattacharya S, et al.

551

Recent advances in 2D and 3D in vitro systems using primary hepatocytes,

552

alternative hepatocyte sources and non-parenchymal liver cells and their

553

use in investigating mechanisms of hepatotoxicity, cell signaling and ADME.

554

Arch Toxicol 2013;87:1315–1530.

555

[4] Schliess F, Hoehme S, Henkel SG, Ghallab A, Driesch D, Bottger J, et al.

556

Integrated metabolic spatial-temporal model for the prediction of ammonia

557

detoxification during liver damage and regeneration. Hepatology

558

2014;60:2040–2051.

559

[5] Hoehme S, Brulport M, Bauer A, Bedawy E, Schormann W, Hermes M, et al.

560

Prediction and validation of cell alignment along microvessels as order

561

principle to restore tissue architecture in liver regeneration. Proc Natl Acad

562

Sci U S A 2010;107:10371–10376.

563

[6] Campos G, Schmidt-Heck W, Ghallab A, Rochlitz K, Putter L, Medinas DB,

564

et al. The transcription factor CHOP, a central component of the

565

transcriptional regulatory network induced upon CCl4 intoxication in

566

mouse liver, is not a critical mediator of hepatotoxicity. Arch Toxicol

567

2014;88:1267–1280.

568

[7] Gebhardt R, Baldysiak-Figiel A, Krugel V, Ueberham E, Gaunitz F.

Hepato-569

cellular expression of glutamine synthetase: an indicator of morphogen

570

actions as master regulators of zonation in adult liver. Prog Histochem

571

Cytochem 2007;41:201–266.

572

[8] Levesque R, Leblanc M, Cardinal J, Teitlebaum J, Skrobik Y, Lebrun M.

573

Haemodialysis for severe hyperammonaemic coma complicating urinary

574

diversions. Nephrol Dial Transplant 1999;14:458–461.

575

[9] Enns GM, Berry SA, Berry GT, Rhead WJ, Brusilow SW, Hamosh A. Survival

576

after treatment with phenylacetate and benzoate for urea-cycle disorders. N

577

Engl J Med 2007;356:2282–2292.

578

[10] Poh Z, Chang PE. A current review of the diagnostic and treatment strategies

579

of hepatic encephalopathy. Int J Hepatol 2012;2012 480309.

580

[11] Clay AS, Hainline BE. Hyperammonemia in the ICU. Chest

581

2007;132:1368–1378.

582

[12] Rajpoot DK, Gargus JJ. Acute hemodialysis for hyperammonemia in small

583

neonates. Pediatr Nephrol 2004;19:390–395.

584

[13] Summar M. Current strategies for the management of neonatal urea cycle

585

disorders. J Pediatrics 2001;138:S30–S39.

586

[14] Honda S, Yamamoto K, Sekizuka M, Oshima Y, Nagai K, Hashimoto G, et al.

587

Successful treatment of severe hyperammonemia using sodium

phenylac-588

etate powder prepared in hospital pharmacy. Biol Pharm Bull

589

2002;25:1244–1246.

590

[15] Misel ML, Gish RG, Patton H, Mendler M. Sodium benzoate for treatment of

591

hepatic encephalopathy. Gastroenterol Hepatol 2013;9:219–227.

592

[16] Collen JF, Das NP, Koff JM, Neff RT, Abbott KC. Hemodialysis for

hyperam-593

monemia associated with ornithine transcarbamylase deficiency. Appl Clin

594

Genet 2008;1:1–5.

595

[17] Boon L, Geerts WJ, Jonker A, Lamers WH, Van Noorden CJ. High protein diet

596

induces pericentral glutamate dehydrogenase and ornithine

aminotrans-597

ferase to provide sufficient glutamate for pericentral detoxification of

598

ammonia in rat liver lobules. Histochem Cell Biol 1999;111:445–452.

JOURNAL OF HEPATOLOGY

(15)

599 [18] Sies H, Häussinger D, Grosskopf M. Mitochondrial nicotinamide nucleotide

600 systems: ammonium chloride responses and associated metabolic

transi-601 tions in hemoglobin-free perfused rat liver. Hoppe Seylers Z Physiol Chem

602 1974;355:305–320.

603 [19] Spanaki C, Plaitakis A. The role of glutamate dehydrogenase in mammalian

604 ammonia metabolism. Neurotox Res 2012;21:117–127.

605 [20] McGill MR, Sharpe MR, Williams CD, Taha M, Curry SC, Jaeschke H. The

606 mechanism underlying acetaminophen-induced hepatotoxicity in humans

607 and mice involves mitochondrial damage and nuclear DNA fragmentation. J

608 Clin Invest 2012;122:1574–1583.

609 [21] Meuret G, Beck K, Keul J, Gruenagel HH. On therapy of hepatic coma with

610 metabolites of the urea cycle, glutamate and alpha-ketoglutarate. Dtsch Med

611 Wochenschr 1968;93:1194–1197.

612 [22] Wildhirt E. The terminal stage of liver diseases and hepatic coma. Der

613 Internist 1965;6:439–446.

614

[23] Gebhardt R, Mecke D. Heterogeneous distribution of glutamine synthetase

615

among rat liver parenchymal cells in situ and in primary culture. EMBO J

616

1983;2:567–570.

617

[24] Häussinger D. Hepatocyte heterogeneity in glutamine and ammonia

618

metabolism and the role of an intercellular glutamine cycle during

619

ureogenesis in perfused rat liver. Eur J Biochem 1983;133:269–275.

620

[25] Stoll B, Häussinger D. Hepatocyte heterogeneity in uptake and metabolism

621

of malate and related dicarboxylates in perfused rat liver. Eur J Biochem

622

1991;195:121–129.

623

[26] Stoll B, McNelly S, Buscher HP, Häussinger D. Functional hepatocyte

624

heterogeneity in glutamate, aspartate and alpha-ketoglutarate uptake: a

625

histoautoradiographical study. Hepatology 1991;13:247–253.

626