Eur. J. Biochem.256, 16223 (1998) FEBS1998

The influence of nickel and cobalt on putative members of the oxygen-sensingpathway of erythropoietin-producing HepG2 cells

Torsten PORWOL1, Wilhelm EHLEBEN1, Karl ZIEROLD1, Joachim FANDREY2 and Helmut ACKER1

1 Max Planck Institut für molekulare Physiologie, Dortmund, Germany2 Institut für Physiologie, Medizinische Universität zu Lübeck, Lübeck, Germany

(Received 29 April1998) 2 EJB 98 0588/1

Cobalt and nickel stimulate, as does hypoxia, the production of erythropoietin (EPO) in HepG2 cells.Under hypoxic conditions, a decrease in the level of intracellular reactive oxygen species (ROS) is thoughtto stimulate EPO expression. Cobalt and nickel may interact with the putative oxygen sensor by changingthe redox state of the central iron atom of heme proteins, similar to the effects of hypoxia. It wasinvestigated, therefore, whether cobalt and nickel interact with hemeproteins or ROS scavenging systemsin the control of intracellular ROS level. Cobalt chloride (100 µM, 24 h) oxidized non respiratory as wellrespiratory hemeproteins and increased the oxygen consumption. In contrast, nickel chloride (300µM,24 h) primarily reduced respiratory hemeproteins and decreased the oxygen consumption. In HepG2 cellstreated with CoCl2, iron and cobalt were localized in cytosolic granules close to the cell nucleus and inmitochondria at concentrations up to12 mM or 41 mM, respectively. Intracellular nickel was not measur-able. Three-dimensional reconstruction of confocal laser microscopy images revealed hot spots of hy-droxyl radical generation by a Fenton reaction at the sites of cytosolic iron accumulation. TheÙOH levelsdecreased in cobalt-treated (to 81%) as well as in nickel-treated (to 67%) HepG2 cells, accompanied byan increase of EPO expression to167% and150%, respectively. Our results underline the importance ofÙOH formed by a Fenton reaction for triggerimg EPO production. Identification of the primary hemeprot-ein being the oxygen sensor was not possible due to the antagonistic effects of cobalt and nickel on theredox state of detectable hemeproteins.

Keywords:heme protein; oxygen consumption; nickel; cobalt; reactive oxygen species.

The involvement of a hemeprotein as part of the cellularoxygen sensor in the control of hypoxia-inducible gene expres-sion has been proposed by Goldberg et al. [1]. The hypothesisis based on the stimulatory effects of cobalt, nickel and irondepletion in mimicking hypoxic induction of erythropoietin(EPO) production in the human hepatoma cell Hep3B. Numer-ous publications followed confirming the importance of thesetransition metals for the analysis of the oxygen-sensing signalpathway. Cobalt chloride and the iron chelator desferrioxaminemimicked venous pO2 and enhanced the insulin-dependent in-duction of glucokinase mRNA in rat hepatocytes despite a highpO2 corresponding to the arterial oxygen tension [2]. Cobalt and/or nickel substituted for the hypoxic stimulus to induce endo-thelin-1 in early passages of human umbilical endothelial cells[3], the human phosphoglycerate kinase1 and mouse lactate de-hydrogenase A genes [4]. Cobalt and desferrioxamine increasedthe expression of mRNAs for vascular endothelial growth factor(VEGF), platelet-derived growth factor (PDGF) A and B chains,placental growth factor (PLGF), transforming growth factor [5]and mRNAs encoding phosphofructokinase, aldolase, lactate de-

Correspondence toH. Acker, Max-Planck-Institut für molekularePhysiologie, Rheinlanddamm 201, D-44139 Dortmund, Germany

Fax: 141231 1206 464/530.E-mail : helmut.acker@mpi-dortmund.mpg.deAbbreviations.EPO, erythropoietin; ROS, reactive oxygen species;

VEGF, vascular endothelial growth factor; PDGF, platelet-derivedgrowth factor; PLGF, placental growth factor; HIF, hypoxia-induciblefactor ; M-MLRV, Moloney Murine Leukemia Virus; IRE, iron respon-sive element ; TfR, transferrin receptor.

hydrogenase and the glucose transporters GLUT-1 and GLUT-3 [6]. Ho and Bunn [7] have put forward the notion that ironand cobalt compete for the heme protein, since Hep3B cells in-cubated in low iron medium required less cobalt for EPO stimu-lation than those exposed to iron-enriched medium. Apart frominteracting with heme proteins, however, transition metal ionscould control the degradation of ROS, possible messengers ofthe oxygen-sensing signal pathway as suggested by Ehleben etal. [8] and Fandrey et al. [9]. Hydrogen peroxide detected inEPO-producing HepG2 cells [10] depressed hypoxia-stimulatedEPO production [11]. The inhibition was antagonized by ironchelators and theÙOH scavengers dimethylthiourea or tetrameth-ylthiourea [9], indicating the involvement of a Fenton-type reac-tion in the action of ROS. Likewise, H2O2 inhibited the insulin-dependent induction of glucokinase mRNA increased under ve-nous pO2 and restored glucagon-dependent phosphoenolpyru-vate carboxykinase mRNA expression that was lowered by ve-nous pO2 [2, 12]. On the molecular level, the effects, at least forROS on EPO production, have been attributed to an increase inthe transcription factor hGATA2, which acts as an repressor atthe EPO promoter in Hep3B cells [13]. In addition, H2O2 isdescribed to affect the redox-sensitive EPO RNA-binding pro-tein [14] and to decrease the stability of the hypoxia-induciblefactor 1A protein (HIF-1A) [15]. HIF-1A and HIF-1β form theheterodimeric transcription factor HIF-1 that is the principal reg-ulator in the hypoxic induction of a number of physiologicallyimportant genes (see for review [16]).

From these investigations we hypothesize the following oxy-gen-sensing signal pathway. A heme protein containing an oxy-

17Porwol et al. (Eur. J. Biochem. 256)

gen sensor forms less ROS under hypoxic conditions, resultingin a decreasedÙOH production by an intracellular Fenton reac-tion. As a consequence, stabilized HIF-1A protein heterodimer-ises with theβ subunit to form the HIF-1 complex and, in con-cert with decreased levels of the repressor GATA2, induces EPOgene expression. An increased EPO mRNA lifetime by upregu-lation of the EPO mRNA-binding protein could further augmentEPO mRNA levels. Herein, the question was addressed whethertransition metals interact with upstream putative members of theoxygen-sensing signal pathway like heme proteins, Fenton reac-tion and ROS. Light absorption spectrophotometry of heme pro-teins, pO2 measurements for oxygen-consumption determina-tions, electron probe X-ray microanalysis for intracellular locali-sation of transition metals, three-dimensional localisation byconfocal laser microscopy of cellularÙOH production as well asdetermination of EPO mRNA levels were used. As a single com-mon action of cobalt and nickel, we found enhanced EPOmRNA levels and decreased intracellularÙOH concentrations.Antagonistic effects of cobalt and nickel as well as a negligibleeffect of iron depletion on the absorption spectra and the redoxstate of hemeproteins [8] make it difficult to identify primehemeproteins as part of the oxygen sensor.

MATERIALS AND METHODS

Tissue culture. HepG2 cells (ATCC HB 8065) were culti-vated in monolayer and spheroid tissue culture as described byGörlach et al. [10] in RPMI 1640 medium (Life Technologies)supplemented with10% fetal calf serum, penicillin (100 U/ml)and streptomycin (100 µg/ml) at 37°C in an incubator contain-ing humid air with 5% CO2 (Stericult 200, Labotect). Spheroidcultures were started with 2.53105 cells in 130 ml medium inagarose-treated petri dishes (Life Technologies). After reachinga diameter of150 µm, spheroids were further cultured in sili-cone-treated cylindrical spinner flasks (11 cm diameter and24 cm height) with a spin rate of 40 rpm containing 250 ml me-dium which was renewed twice/week. For incubation with co-balt chloride (100 µM) or nickel chloride (300µM), spheroidswith a diameter of 7002800µm were kept in petri dishes (LifeTechnologies) for 24 h. The duration of incubation as well asthe concentration of cobalt and nickel were shown by Goldberget al. [1] as optimal for stimulation of EPO production in Hep3Bcells.

Light absorption photometry. Spheroids (diameter range7002800µm) were located in a superfusion chamber on a smallopaque bench with little holes of the same diameter as the spher-oids. Superfusion was performed as described [10, 17]. Briefly,isotonic salt solution (Locke’s solution) containing glucose(5 mM) was equilibrated with different O2/CO2 mixtures in orderto adjust the oxygen tension to various levels at pH 7.42. Theflow rate through the chamber was10 ml/min. Spheroids weresupplied symmetrically with nutrients by this procedure. Thetemperature was maintained at 36°C. The oxygen tension, pHand temperature in Locke’s solution were controlled by corre-sponding electrodes. The superfusion chamber was mounted onthe stage of a light microscope (Olympus) for light absorptionmeasurements. White light from a halogen bulb (12 V, 100 W)transilluminating the spheroid only passed to an objective (403)and was analyzed for absorbance changes at different wave-lengths by a photodiode-array spectrophotometer (MCS 210,Zeiss) connected to the third ocular of the microscope trinocularhead via a light guide. Difference spectra were obtained byfirstly recording in superfusion medium equilibrated with 20%O2, 3% CO2 and 77% N2 (aerobic steady state) which was auto-matically substracted from the spectra under reducing conditions

by equilibrating with 3% CO2 and 97% N2. The recorded differ-ence spectra were evaluated, deconvoluted and visualized usingthe software package TechPlot (Dr Dittrich, Braunschweig, Ger-many).

pO2 measurements.Glas-isolated platinum electrodes witha tip diameter of 224 µm as described by Baumgärtl andLübbers [18] were used for polarographic pO2 measurementsin HepG2 spheroids. For calibration, pO2 microelectrodes wereintroduced into the medium flowing through the above describedsuperfusion chamber by means of the micromanipulator 5171from Eppendorf. The pO2 in the medium was then changed byequilibrating the medium with different gas mixtures containing0%, 10% or 20% O2 in 5% CO2 the remaining gas was N2.Calibration curves were constructed from the polarographic re-duction current of the electrode and the different pO2 values inthe medium before and after each measurement. Only resultsfrom stable electrodes were considered [8,19]. The pO2 micro-electrode positioned on the upper middle point of the spheroidswas moved into the spheroid tissue in 50-µm steps perpendicularto the vertical axis for tissue pO2 measurements.

Oxygen consumption (VO2, expressed as ml O2 · 100 g wettissue21 · min21) was calculated from the measured pO2 profileinside the HepG2 spheroids [19] by adapting Henry’s law:

pO2(r) 5 pO2(s) 2 VO2 (6DA)21 (R2 2 r2)

where pO2(r) 5 pO2 at r, pO2(s) 5 pO2 at the surface, D5oxygen diffusion coefficient for tumor tissue,A 5 oxygen solu-bility coefficient for tumor tissue [20],R 5 radius of the spher-oid, r 5 distance to the center.

Radical oxygen species and confocal laser microscopy.For measuring intracellular ROS levels under the influence ofcobalt and nickel HepG2 cells cultured in 96 multi-well plateswere stained with dihydrorhodamine123 at a concentration of25 µM (Molecular Probes) for 30 min [10]. Dihydrorhodamine123 is a non-fluorescent agent described as specific for detectingH2O2 (Molecular Probes). In the presence of iron, H2O2 in aFenton-type reaction rapidly and specificaly oxidizes dihydro-rhodamine123 to rhodamine123 which emits fluorescent lightwhen excited [21]. One important intermediate in the Fenton-type reaction isÙOH which has been detected by several spectro-scopical methods [22] and is also ultimately responsible for theconversion of non-fluorescent dihydrorhodamine123 to fluores-cent rhodamine123 [8]. Intracellular rhodamine123 fluores-cence intensity was measured after careful washing with 2.7 mMKCl, 1.5 mM KH2pO4, 136.9 mM NaCl, 8.1 mM Na2HPO4,5 mM CaCl2, pH 7.4 (NaCl/Pi) in each well after 30 min usinga multiwell scanning fluorescence photometer (Titertek, Fluo-reskan II, Flow). 48 wells were recorded in each experiment.

For three-dimensional visualization of theÙOH distributioninside HepG2 cells, optical sections were performed as de-scribed by Ehleben et al. [8]. Briefly, a confocal laser scanningmicroscope (MRC 600, Bio-Rad Inc.) was attached to an in-verted microscope (ICM 405, Zeiss) having an 603objective(CF N Plan Apochromat, Nikon GmbH) with cover slip correc-tion, a numerical aperture of1.20 and a working distance of220µm. 39242 optical sections with a resolution of1923256/channel were recorded with a step size of1 µm using the SOMsoftware (Bio-Rad) and stored in the native PIC file format (ver-sion 2.0). Typically, two pictures were averaged in order to mini-mize bleaching of the specimen. In order to visualize theÙOHproduction in relation to the cell border, we used SNARF calceinAM (70 µM) which is described as applicable for intracellularpH measurements (Molecular Probes). We selected SNARF asa second dye, because it displays minimal interference with theÙOH labelling. The two dyes rhodamine123 and SNARF calceinAM were excited with the 488-nm line of an Ar ion laser (Om-

18 Porwol et al. (Eur. J. Biochem. 256)

nichrome, Laser 2000) and the fluorescence was detected usinga band pass filters of 570 nm and 640 nm, respectively. Withinthe focal plane of the microscope, the power of the 488-nm linewas reduced below1mW, the pinhole adjusted to1.9 mm, andthe photo multiplier gain optimized for each individual channel.An electronic zoom of1.522 was used. The sequential 8-bitBio-Rad PIC files were transferred to a UNIX-based workstationcluster via ethernet for image processing and data visualization.The processed data sets and the real cell border coordinateswhich were calculated according to the method of Visser et al.[23] were stored in AVS field file format (Application Visualiza-tion System, AVS Inc.) for further processing on the SPARC10or INDY (Silicon Graphics Inc., SGI) equipped with192 MBRAM. We have developed programs on the basis of AVS inorder to perform the data visualization.

Electron probe X-ray microanalysis. Preparation andanalysis of cryosectioned cells for determining the intracellularion content was done as described previously [24, 25]. HepG2cells grown on Petriperm foil (Heraeus) were cryofixed quicklyby plunging into liquid propane cooled by liquid nitrogen. Ap-proximately 100-nm thick cryosections were prepared in aReichert FC4/Ultracut cryoultramicrotome (Reichert) by a dryglass knife at a temperature below150 K. The cryosections wereplaced on Pioloform-coated and carbon-evaporated Cu grids andtransferred into the cold stage of a Siemens ST100F scanningtransmission electron microscope (Siemens). The sections werefreeze-dried during cryotransfer. The freeze-dried cryosectionswere imaged and analyzed by scanning the electron beam of100 kV and 1.3 nA across areas of interest. X-rays were col-lected by means of a Si(Li) detector with Be window (NuclearSemiconductor) and analyzed with respect to their energy by aLink AN 10 000 multichannel analyzer (Link). From the X-rayspectra element contents related to dry mass were calculated byuse of the Link Quantem FLS program based on the Hall contin-uum method.

Erythropoietin determination. HepG2 spheroids treatedeither with cobalt or nickel for 24 h were washed with NaCl/Pi

after removal of the spent cell culture medium and lysed with700µl guanidinium thiocyanate (4 M with 0.1 M 2-mercapto-ethanol). Total RNA was extracted by the acid/phenol method[26]. After dissolving the RNA in diethylpyrocarbonate-treatedwater, the concentration was determined by measuring the absor-bance at 260 nm. To check the integrity of the RNA, aliquotswere run on a1.1% formaldehyd/agarose gel.1 µg total RNAwas reverse transcribed into first-strand cDNA using oligodT15

as primer for the Moloney Murine Leukemia Virus ReverseTranscriptase (M-MLRV RT) superscript (GIBCO Life Technol-ogies) in a total volume of 25µl. Reverse transcription was per-formed at 42°C for 60 min after an initial denaturation step at68°C for 10 min. The reaction was terminated by boiling thesamples for10 min. Until quantification by a competitive PCRcDNA stocks were kept at220°C. Competitive PCR for EPOmRNA was performed exactly as described [27].

Statistical analysis.Data are presented as mean6 standarddeviations. For statistical analyses, the Student’st-test was per-formed. Differences were considered significant when thePvalue was,0.05.

RESULTS

Absorption photometry of cytochromes. Mean values of N2versus aerobic steady-state difference spectra (solid line, controlconditions) in16 single HepG2 spheroids shown (Fig.1a) twolight absorption peaks with band maxima at 550 nm and 602 nmas well as a broader shoulder between 558 nm and 563 nm.

Fig. 1. N2 versus aerobic steady-state difference spectra (solid line).Values are the means of16 experiments carried out under control condi-tions. (a); (b) eight experiments with 24 h treatment with100 µM CoCl2;(c) seven experiments with 24 h treatment with 300µM NiCl2 (c). Thebasic redox spectra of isolated cytochromec peaking at 550 nm (5),cytochromeb558 (\\), cytchromeb563 (11) and cytochromeaa3 peakingat 605 nm (||) are shown. Light absorbance changes of the spectra of theisolated cytochromes are relative values that were calculated to fit theexperimental curve (solid line) as close as possible by a superpositioncurve (open circles).

Since this spectrum is composed of different mitochondrial andnon-mitochondrial cytochromes, difference spectra of variousisolated cytochromes have been used to identifiy the peaks andthe shoulder as well as to fit the experimental curve by asuperposition curve. The curves (Fig.1a) marked with differentsymbols correspond to redox difference spectra of isolated mito-chondrial cytochromec peaking at 550 nm,b563 peaking at563 nm andaa3 peaking at 605 nm [28, 29] as well as to thenon-mitochondrial cytochromeb558 peaking at 558 nm as de-scribed for the NADPH oxidase in neutrophils [30]. The ampli-tude of the absorbance for the spectra of the isolated cyto-chromes was varied to fit the experimental curve (Fig.1a, solidline) as close as possible with superposition curve (open circles).The peaks at 550 nm and 602 nm are clearly related to cyto-chromec and cytochromeaa3. The peak of isolated cytochromeaa3 is shifted to 605 nm due to its reduction by cyanide [29].This shift of the peak wavelength between hypoxia and cyanide-

19Porwol et al. (Eur. J. Biochem. 256)

Fig

.2.

The

intr

acel

lula

rco

nver

sion

ofno

n-flu

ores

cent

dihy

dror

hoda

min

e12

3to

fluor

esce

ntrh

odam

ine

123

med

iate

dby

.O

Hin

dica

ting

aF

ento

nre

actio

nas

visu

aliz

edth

ree-

dim

ensi

onal

lyfo

ron

efu

llH

epG

2ce

llan

dpa

rts

ofth

ree

othe

rs.The

cell

bord

ers

are

show

nby

isol

ines

.T

herh

odam

ine

123

fluor

esce

nce

leve

ls22

0,20

0an

d150

are

show

nse

para

tely

inth

ree

diffe

rent

pane

lsw

itha

colo

urco

dera

ngin

gfr

ombl

ueto

red

whi

cheq

uals

ara

nge

for

the

arbi

trar

yun

itsof

02

255.

The

thre

epa

nels

show

the

Hep

G2

cells

from

diffe

rent

angl

esin

orde

rto

enha

nce

the

thre

e-di

men

sion

alim

pres

sion

ofce

llex

tens

ions

and

loca

lisat

ion

ofF

ento

nre

actio

n.T

hepa

nela

tth

elo

wer

right

corr

elat

esth

edi

ffere

ntrh

odam

ine

123

fluor

esce

nce

inte

nsity

leve

lsdr

awn

atth

ex-

axis

toth

era

tioof

the

volu

me

occu

pied

byth

erh

odam

ine1

23flu

ores

cenc

ein

tens

ityve

rsus

the

tota

lcel

lvol

ume

draw

nat

the

y-ax

is.

reduced cytochromeaa3 could also be observed in our experi-ments on HepG2 spheroids (data not shown). The shoulderseems to be composed of cytochromeb563 and cytochromeb558.Deconvolution of the spectrum (Fig.1a) allows the maxima ofthe different cytochromes to be related to each other and to assesthe influence of nickel and cobalt on this relationship, which iseither indicative of a change in the redox state and/or the totalamount of the different cytochromes. This assessment is not in-fluenced by the shift of the peak absorption wavelength of cyto-

chromeaa3 since one can mathematically compensate the shiftfrom 605 nm to 602 nm (data not shown). Under control condi-tions, cytochromesc, b558, b563 and aa3 are related as1:0.04 :0.33:0.99. The ratio indicates that the degree of reduc-tion by N2 superfusion depends on the oxidation state of thecytochromes under 20% O2 superfusion. Cytochromeaa3 nor-malized to cytochromec is oxidized to the highest degree, asone would expect from the different redox potentials of the re-spiratory chain cytochromes [28]. The deconvolution of a N2

20 Porwol et al. (Eur. J. Biochem. 256)

Fig. 3. Freeze-dried cryosection of a HepG2 cell under control conditions (a) and after 24 h incubation with 100µM CoCl 2 (c). The spots ofanalysis by electron-probe X-ray microanalysis are marked by an arrow and can directly be related to the X-ray spectra, (b, d) indicating theintracellular concentration of different ions. n5 nucleus, m5 mitochondrium, c5 cytoplasm. The cell dimensions are given by bars.

versus aerobic steady-state difference spectrum after 24 h treat-ment with 100 µM CoCl2 shows a relationship of1 :0.2:0.44 :0.91, calculated from eight spheroids and confirm-ing deconvolution data as reported by Ehleben et al. [8]. After24 h treatment with 300µM NiCl2 (n 5 7) the cytochromes ofthe N2 versus aerobic steady-state difference spectra are relatedas1 :0.03:0.2:0.8 (Fig.1c).

pO2 gradients and oxygen consumption:Measurements ofpO2 gradients by microelectrodes in HepG2 spheroids, as de-scribed by Görlach et al. [10], were used to assess whether theeffect of nickel on the difference spectra could be corroboratedby changes in the oxygen consumption rate. Spheroids (radius372626 µm, n 5 10) treated with 300µM NiCl2 for 24 h havesignificant flatter pO2 gradients than the control spheroids (ra-dius 407639 µm, n 5 8). The lowest, central pO2 value in con-trol spheroids amounted to 665 mm Hg and in nickel-treatedspheroids significantly higher to1667 mm Hg (P<0.023). Cal-culated from these values the oxygen consumption in controlspheroids was 0.65 ml O2/100 g min and, in nickel-treated spher-

oids, 0.57 ml O2/100 g min. This reduced oxygen consumptionindicates an reducing effect of nickel on the respiratory cyto-chromes. In contrast, oxygen consumption rate of HepG2 spher-oids increased after 24 h treatment with100 µM CoCl2 as pre-viously reported [8,17] hinting to an oxidizing effect of cobalton the respiratory cytochromes.

Reactive oxygen species and confocal laser microscopy.The intracellular conversion of nonfluorescent dihydrorhoda-mine 123 to fluorescent rhodamine123 mediated byÙOH(Fig. 2) show the outer surface of the cells by isolines. The in-tensity level of the rhodamine123 fluorescence ranges fromover 02255 arbitrary units, equivalent to the colour-code rangefrom blue to red. The fluorescence levels of 220, 200 and150are shown separately in three panels. Using these data, the vol-ume which is occupied by rhodamine123 fluorescence and thetotal cell volume represented by the isolines can be calculated[31]. The result of this calculation (Fig. 2) correlate the differentfluorescence levels at 02255 on thex-axis with the ratio of thecell volume occupied by the rhodamine123 fluorescence versusthe total cell volume on they-axis. In agreement with the cell

21Porwol et al. (Eur. J. Biochem. 256)

Table 1. Control Hep G2 cells cultured on Petriperm foil.n 5 numberof measurements, n.s.5 not significant.

Element Cytoplasm Nucleus Mitochondria

mmol · kg H2O216SD

Na 206 16 106 91 46 9Mg 126 4 86 4 266 7P 1816 28 1656 20 187654S 556 14 496 13 92615Cl 436 14 326 5 126 4K 1766 29 1756 20 169646Ca n.s. n.s. n.s.Fe 56 4 n.s. 46 3Co n.s. n.s. n.s.

n 10 10 10

Table 2. Hep G2 cells cultured on Petriperm foil after incubationin 100 µmol/l CoCl2 for 24 h. n 5 number of measurements, n.s., notsignificant.

Element Cytoplasm Cytoplasmic Mitochondria Mitochondrialgranules granules

mmol · kg H2O216SD

Na 66662 166 10 13612 96 18Mg 106 4 166 20 17610 96 21P 110627 4226 108 272680 2846 148S 44610 1656 58 90623 2146 77Cl 59626 456 26 136 6 576 36K 103649 1886 62 95646 1096 62Ca n.s. n.s. n.s. n.s.Fe 36 1 126 8 46 2 126 5Co 56 1 356 12 96 6 416 15

n 10 5 10 8

pictures, high rhodamine123 fluorescence levels are correlatedwith low ratio values, meaning that a high degradation rate ofH2O2 by the Fenton reaction takes place only at few hot spotswithin the cells. Much lower levels either indicating low degra-dation rates or that diffusion of rhodamine123 from the few hotspots into the cytosol are found at other places with more than50% of the cell and the nucleus free of rhodamine123 fluores-cence.

Treatment of HepG2 cells with cobalt decreases the meanintracellularÙOH level to 8169% of control (n 5 5, P,0.0028)and with nickel to 6769% (n 5 5, P,0.0003).

Intracellular ion content. For freeze-dried cryosections ofHepG2 cells (Fig. 3), the spots of analysis by electron-probe X-ray microanalysis are marked by an arrow, and correspondingX-ray spectra (b,d) show the intracellular concentration of dif-ferent ions. Iron was measured in both cytoplasm and mito-chondria. The results, together with other ion concentrations un-der control conditions, are summarized in Table1. The concen-tration of calcium was below the detection limit of the methodapplied. HepG2 cells (n 5 10) incubated with 300µM NiCl 2

for 24 h revealed intracellular ion concentrations statistically notdifferent from the control values (data not shown). Nickel, how-ever, was not detectable. In contrast, cobalt was found intracellu-larly, in part agglomerated in cytosolic or mitochondrial gran-ules. Table 2 summarizes the ion concentrations detected at dif-ferent intracellular locations in HepG2 cells treated with100µM

CoCl2 for 24 h, which induced the formation of cytoplasmic andmitochondrial granules that contained cobalt and other ions. Co-balt was also found in the cytoplasm and in mitochondria, indi-cating an active uptake of this transiton metal by HepG2 cells.The somewhat elevated cytoplasmic concentrations of sodiumand chloride together with the decreased potassium level mightindicate a beginning membrane leakage.

EPO production. HepG2 spheroids with a diameter of 7002800µm were treated in two experiments with cobalt chloride(100 µM) or nickel chloride (300µM) for 24 h. Each experimentincludes10 spheroids. EPO mRNA increased to167% in cobalt-treated and150% in nickel-treated spheroids. These results indi-cate that HepG2 cells in multicellular spheroid culture as theywere used for photometric studies respond to transition metalswith an increased EPO production however to a lower degreethan known from HepG2 cells in monolayer culture [9]. Thismight be due to low pO2 values in the central regions of thespheroids which also diminished the magnitude of the hypoxia-stimulated EPO production in HepG2 spheroids [10, 17].

DISCUSSION

The present investigation has shown that the transition met-als cobalt and nickel lead to a decrease in the intracellularÙOHlevel of HepG2 cells with a concomitant increase of the EPOproduction. This is in line with the effect of desferrioxamine onthe ÙOH level and EPO production of HepG2 cells [8, 9]. Itfurthermore confirms the potential role of ROS as intracellularmessenger molecules in control of gene transcription, cytokine,growth factor and hormone production and action, ion transport,neuromodulation and apoptosis [32]. But how do transition met-als affect the intracellularÙOH level?

Significant intracellular concentrations of ROS, as H2O2, sin-glet oxygen (1O2), superoxide anion (O22 ) and ÙOH are thoughtto be produced during normal aerobic metabolism [33]. H2O2,which is a non-charged molecule, can cross membranes and par-ticipate in one-electron transfer reactions. Although it is a suit-able candidate for an intracellular messenger, H2O2 is enzymati-cally degraded by cytosolic and mitochondrial glutathione per-oxidase, which can be enhanced by hypoxia and cobalt as re-cently shown for HepG2 cells [8]. The cytosolic and mito-chondrial localization of cobalt (Fig. 3) matches the localizationof glutathione peroxidase. Nickel, however, seems to lower theglutathione peroxidase activity in liver cells without any distincteffect on the catalase activity the second important enzymaticscavenger [34].

H2O2 is a relatively inert reaction partner. The detection ofintracellular hot spots of highÙOH generation (Fig. 2) and irondeposits (Fig. 3) strongly suggest that H2O2 is degraded in a Fen-ton-type reaction very locally. Fenton-type reactions need transi-tion metals such as iron, cobalt and nickel [35]. The iron-medi-ated Fenton reaction gives rise toÙOH as the most importantradical produced [36]. In contrast, hydroperoxy radical (ÙHO2) isthe predominant radical produced by the cobalt-mediated Fentonreaction as shown by ESR measurements.In vitro spin trappingexperiments have shown that Ni21 in the presence of H2O2 gen-erates predominantly alkyl and alkoxyl radicals in contrast toNi31 generatingÙOH [37]. Furthermore, nickel strongly inhibitsthe superoxide dismutase activityin vitro [38]. Therefore, thedecrease in the intracellularÙOH level under cobalt and nickeltreatment might be caused by a decreased supply of H2O2 tothe Fenton reaction due to an increased enzymatic activity ofscavenging systems or by a conversion of the Fenton reaction.The importance of the intracellular iron stores for the Fenton-

22 Porwol et al. (Eur. J. Biochem. 256)

reaction-mediatedÙOH production is also substantiated by thefact that inhibition of EPO production by H2O2 can be overcomeby iron chelators [8, 9]. Desferrioxamine, which can diffuse eas-ily into cells where it binds iron in the nonheme labile intracellu-lar pool [39], decreases intracellularÙOH levels by inhibiting theFenton reaction [8].

Iron is involved in an oxygen-sensing system in bacteria andpossibly in mammalian cells as well since iron sulfur (Fe-S)clusters which regulate the activity of transcription factors. Thetranscription factor FNR fromEscherichia coliregulates tran-scription of genes in response to oxygen deprivation due to aFe-S cluster in the wild-type protein that is extremely oxygenlabile. Anoxically purified FNR was mostly a dimer with highDNA-binding activity compared to that of the aerobically puri-fied FNR monomer [40]. The protein SoxR, a homodimer con-taining two (2Fe-2 S) clusters, is both a sensor of oxidativestress and a transcriptional regulator. The purified protein is ahomodimer. One-electron oxidation of the SoxR (2Fe-2 S) clus-ters is the mechanism both for sensing oxidative stress and foractivating the SoxR protein as a transcription factor [41]. Theinactivation-reactivation of aconitase in cultured mammaliancells depends on oxygen radicals with an aconitase 50% activeat a O2

2 concentration of 502200 pM and 86% active at a O22

concentration of 8230 pM [42]. The active RNA-binding formof the aconitase without 4Fe-4 S cluster binds to the 3′ iron re-sponsive element of the transferrin receptor (TfR) mRNA andincreases its stability. TfR expression and subsequently iron up-take increase. In contrast, inactive aconitase with 4Fe-4 S clusterdecreases TfR stability and can sequently lower uptake of ironinto cells [39]. Studies of the influence of transition metals onthe Fe-S-containing transcription factors might be useful to com-plete our understanding the interaction of cobalt and nickel withthe oxygen-sensing pathway.

Transition metals as well as iron chelators have been hypoth-esized to influence the redox state of a putative hemeproteinoxygen sensor upstream in the signal cascade for regulating EPOproduction [1]. The photometric studies as (Fig.1) revealed dif-ferent mitochondrial and non mitochondrial cytochromes inHepG2 cells in accordance with data from the literature [8,10,17]. Obviously, basic spectra of all these cytochromes were nec-essary to fit the experimental difference spectra to assess theeffect due to cobalt and nickel application. An oxidizing effectof cobalt on cytochromeb558 and cytochromeb563 and a generalreducing effect of nickel on the mitochondrial cytochromes asindicated by the deconvolution method could be confirmed bythe oxygen-consumption measurements. Probably in line withthis explanation is the mitochondrial agglomeration of cobalt(Fig. 3). Since reduction of the respiratory chain in HepG2 cellsby cyanide leads to a decrease of the intracellularÙOH level [8],nickel might as well act in the same way. However, a reductionof the respiratory chain by cyanide does not stimulate EPO pro-duction as does nickel [43], and Hela cells remain fully respon-sive to hypoxic stimulation under cyanide poisioning [44]. Sincecobalt and nickel interact with mitochondrial and non-mito-chondrial cytchromes at the same time, it does not yet appear tobe justified to point to one specific hemeprotein as the cellularoxygen sensor. In this respect, results from experiments withiron chelators do not help us any further either [8]. This is inline with desferrioxamine and cobalt affecting tyrosine hydrox-ylase gene expression, being inversely correlated with H2O2 for-mation in PC12 cells, in the mechanism downstream from theH2O2 formation rather than by interfering with the H2O2-generat-ing activity of the oxygen sensor [45].

Transition metals and iron chelators may act at different lev-els of the oxygen-sensing signal pathway, resulting in a decreasein the intracellularÙOH level. This would stabilize HIF-1 and

enhance EPO production. Further experimental approaches arenecessary, however, to discriminate and identify single membersof this pathway as well as to show that transition metals byinteracting with the oxygen-sensing pathway mimic hypoxia.

The technical assistance by B. Bölling and E. Merten is gratefullyacknowledged. This work was financially supported by the DFG grantsAc 37/9-1, Ac 37/9-2.

REFERENCES

1. Goldberg, M. A., Dunning, S. P. & Bunn, H. F. (1988) Regulationof the erythropoietin gene: evidence that the oxygen sensor is aheme protein,Science 242, 141221415.

2. Kietzmann, T., Roth, U., Freimann, S. & Jungermann, K. (1997)Arterial oxygen partial pressures reduce the insulin-dependent in-duction of the perivenously located glucokinase in rat hepatocytecultures2 mimicry of arterial oxygen pressures by H2O2, Bio-chem. J. 321, 17220.

3. Bodi, I., Bishopric, N. H., Discher, D. J., Wu, X. S. & Webster, K.A. (1995) Cell-specificity and signaling pathway of endothelin-1gene regulation by hypoxia,Cardiovasc. Res. 30, 9752984.

4. Firth, J. D., Ebert, B. L., Pugh, C. W. & Ratcliffe, P. J. (1994)Oxygen-regulated control elements in the phosphoglycerate ki-nase1 and lactate dehydrogenase A genes : Similarities with theerythropoietin 3′ enhancer,Proc. Natl Acad. Sci. USA 91, 649626500.

5. Gleadle, J. M., Ebert, B. L., Firth, J. D. & Ratcliffe, P. J. (1995)Regulation of angiogenic growth factor expression by hypoxia,transition metals, and chelating agents,Am. J. Physiol. 268,C13622C1368.

6. Ebert, B. L., Gleadle, J. M., O’Rourke, J. F., Bartlett, S. M., Poulton,J. & Ratcliffe, P. J. (1996) Isoenzyme-specific regulation of genesinvolved in energy metabolism by hypoxia: Similarities with theregulation of erythropoietin,Biochem. J. 313, 8092814.

7. Ho, V. T. & Bunn, H. F. (1996) Effects of transition-metals on theexpression of the erythropoietin gene2 further evidence that theoxygen sensor is a heme protein,Biochem. Biophys. Res. Com-mun. 223, 1752180.

8. Ehleben, W., Porwol, T., Fandrey, J., Kummer, W. & Acker, H.(1997) Cobalt and desferrioxamine reveal crucial members of theoxygen-sensing pathway in HepG2 cells,Kidney Int. 51, 4832491.

9. Fandrey, J., Frede, S., Ehleben, W., Porwol, T., Acker, H. & Jelk-mann, W. (1997) Cobalt chloride and desferrioxamine antagonizethe inhibition of erythropoietin production by reactive oxygenspecies,Kidney Int. 51, 4922496.

10. Görlach, A., Holtermann, G., Jelkmann, W., Hancock, J. T., Jones,S. A., Jones, O. T. G. & Acker, H. (1993) Photometric characteris-tics of haem proteins in erythropoietin producing hepatoma cells(HepG2),Biochem. J. 290, 7712776.

11. Fandrey, J., Frede, S. & Jelkmann, W. (1994) Role of hydrogen-peroxide in hypoxia-induced erythropoietin production,Biochem.J. 303, 5072510.

12. Kietzmann, T., Freimann, S., Bratke, J. & Jungermann, K. (1996)Regulation of the gluconeogenic phosphoenolpyruvate carboxyki-nase and the glycolytic aldolase A gene expression by O2 in rathepatocyte cultures. Involvement of hydrogen peroxide as media-tor in the reponse to O2, FEBS Lett. 388, 2282232.

13. Imagawa, S., Yamamoto, M., Ueda, M. & Miura, Y. (1996) Erythro-poietin gene-expression by hydrogen-peroxide,Inter. J. Hemat.64, 1892195.

14. Rondon, I. J., Scandurro, A. B., Wilson, R. B. & Beckman, B. S.(1995) Changes in redox affect the activity of erythropoietinRNA-binding protein,FEBS Lett. 359, 2672270.

15. Huang, L. E., Arany, Z., Livingston, D. M. & Bunn, H. F. (1996)Activation of hypoxia-inducible transcription factor depends pri-marily upon redox-sensitive stabilization of its a subunit,J. Biol.Chem. 271, 32253232259.

16. Wenger, R. H. & Gassmann, M. (1997) Oxygen(es) and the hypoxia-inducible factor-1, Biol. Chem. 378, 6092616.

23Porwol et al. (Eur. J. Biochem. 256)

17. Görlach, A., Fandrey, J., Holtermann, G. & Acker, H. (1994) Effectsof cobalt on haem proteins of erythropoietin-producing HepG2cells in multicellular spheroid culture,FEBS Lett. 348, 2162218.

18. Baumgärtel, H. & Lübbers, D. W. (1983) Microaxial needle sensorfor polarographic measurement of local O2 pressure in the cellularrange of living tissue. Its contruction and properties, inPolaro-graphic oxygen sensors(Gnaiger, E. & Forstner, H., eds) pp. 37265, Springer-Verlag, Berlin Heidelberg.

19. Gassmann, M., Fandrey, J., Bichet, S., Wartenberg, M., Marti, H.H., Bauer, C., Wenger, R. H. & Acker, H. (1996) Oxygen supplyand oxygen-dependent gene expression in differentiating embry-onic stem cells,Proc. Natl Acad. Sci. USA 93, 286722872.

20. Grote, J., Süsskind, R. & Vaupel, P. (1977) Oxygen diffusivity intumor tissue CDS carcinosarcoma under temperature conditionswithin the range of 20°240°C, Pflügers. Arch. 372, 37242.

21. Royall, J. & Ischiropoulos, H. (1993) Evaluation of 2′,7′-dichloro-fluorescin and dihydrorhodamine123 as fluorescent probes forintracelllular H2O2 in cultured endothelial cells,Arch. Biochem.Biophys. 302, 3482355.

22. Walling, Ch. (1975) Fenton’s reagent revisited,Accounts Chem. Res.8, 1252131.

23. Visser, T. D., Oud, J. L. & Brakenhoff, G. J. (1992) Refractive indexand distance measurements in 3-D microscopy,Optik 90, 17219.

24. Zierold, K. (1988) X-ray microanalysis of freeze-dried and frozen-hydrated cryosections,J. Electron Microsc. Techn. 9, 65282.

25. Zierold, K. (1997) Effects of cadmium on electrolyte ions in cul-tured rat hepatocytes studied by X-ray microanalysis of cryosec-tions,Toxicol. Appl. Pharmacol. 144, 70276.

26. Chomczinski, P. & Sacchi, N. (1987) single step method of RNAisolation by acid guanidiniumthiocyanat-phenol-chloroform-ex-traction,Anal. Biochem. 162, 1562159.

27. Fandrey, J. & Bunn, H. F. (1993) In vivo and in vitro regulation oferythropoietin mRNA, Measurement by competitive polymerasechain reaction,Blood 81, 6172623.

28. Chance, B., Lagallais, V., Sorge, J. & Graham, N. (1975) A versatiletime-sharing multichannel spectrophotometer, reflectometer andfluorometer,Anal. Biochem. 66, 4982514.

29. Heinrich, U. (1981) Untersuchungen zur quantitativen photome-trischen Analyse der Redox-Zustände der Atmungskette in vitround in vivo am Beispiel des Gehirns, PhD Dissertation, Ruhr-Universität Bochum, Germany.

30. Thelen, M., Dewald, B. & Baggiolini, M. (1993) Neutrophil signaltransduction and activation of the respiratory burst,Physiol. Rev.73, 7972821.

31. Thalmann, N. & Thalmann, D. (1991) New trends in animation andvisualization, John Wiley & Sons Inc., Chichester.

32. Lander, H. M. (1997) An essential role for free radicals and derivedspecies in signal transduction,FASEB J. 11, 1182124.

33. Gutteridge, J. M. (1994) Biological origin of free radicals and mech-anisms of antioxidant protection,Chem.-Biol. Interact. 91, 1332140.

34. Misra, M., Rodriguez, R. E. & Kasprzak, K. S. (1990) Nickel in-duced lipid peroxidation in the rat, correlation with nickel effecton antioxidant defense systems,Toxicol. 64, 1217.

35. Lloyd, D. R., Phillips, D. H. & Carmichael, P. L. (1997) Generationof putative intrastrand cross-links and strand breaks in dna bytransition-metal ion-mediated oxygen radical attack,Chem. Res.Toxicol. 10, 3932400.

36. Kadiiska, M. B., Maples, K. R. & Mason, R. P. (1989) A comparisonof cobalt(II) and iron (II) hydroxyl and superoxide free radicalformation,Arch. Biochem. Biophys. 275, 982111.

37. O’Brien, P. & Salacinski, H. J. (1996) Evidence for a relationshipbetween the generation of reactive intermediates and the physico-chemical characteristics of nickel oxides,Arch. Toxicol. 70, 7872800.

38. Shainkin-Kestenbaum, R., Caruso, C. & Berlyne, G. M. (1991) Ef-fect of nickel on oxygen free radical metabolism. Inhibition ofsuperoxide dismutase and enhancement of hydroxydopamineautoxidation,Biol. Trace Element Res. 28, 2132221.

39. Richardson, D. R. & Ponka, P. (1997) The molecular mechanismsof the metabolism and transport of iron in normal and neoplastic-cells,Biochim. Biophys. Acta-Rev. Biomem. 1331, 1240.

40. Lazazzera, B. A., Beinert, H., Khoroshilova, N., Kennedy, M. C. &Kiley, P. J. (1996) DNA binding and dimerization of the Fe-S-containing FNR protein from Escherichia coli are regulated byoxygen,J. Biol. Chem. 271, 276222768.

41. Hidalgo, E., Ding, H. G. & Demple, B. (1997) Redox signal-trans-duction2 mutations shifting [2fe-2s] centers of the soxr sensor-regulator to the oxidized form,Cell 88, 1212129.

42. Gardner, P. R., Raineri, I., Epstein, L. P. & White, C. W. (1995)Superoxide radical and iron modulate aconitase activity in mam-malian cells,J. Biol. Chem. 270, 13399213405.

43. Tan, C. C. & Ratcliffe, P. J. (1991) Effect of inhibitors of oxidativephosphorylation on erythropoietin mRNA in isolated perfused ratkidneys,Am. J. Physiol. 261, 9822986.

44. Jiang, B. H., Semenza, G. L., Bauer, C. & Marti, H. H. (1996)Hypoxia inducible factor1 levels vary exponentially over a physi-ologically relevant range of O2 tension,Am. J. Physiol. 271,C117221180.

45. Kroll, S. L. & Czyzyk-Krzeska, M. F. (1998) Role of H2O2 andheme-containing O2 sensors in hypoxic regulation of tyrosine-hy-droxylase gene-expression,Am. J. Physiol. 43, C1672174.