Relevance oftheincubation period incytotoxicity testing withprimary
human hepatocytes
XiaolongGu
1,2
· WiebkeAlbrecht
1
· KarolinaEdlund
1
· FranziskaKappenberg
3
· JörgRahnenführer
3
·
MarcelLeist
4
· WolfgangMoritz
5
· PatricioGodoy
1
· CristinaCadenas
1
· RosemarieMarchan
1
· TimBrecklinghaus
1
·
LaiaTolosaPardo
6
· JoséV.Castell
6
· IainGardner
7
· BoHan
2
· JanG.Hengstler
1
· ReginaStoeber
1
Abstract
Primary human hepatocytes (PHHs) remain the gold standard for invitro testing in the field of pharmacology and toxicol-
ogy. One crucial parameter influencing the results of invitro tests is the incubation period with test compounds. It has been
suggested that longer incubation periods may be critical for the prediction of repeated dose toxicity. However, a study that
systematically analyzes the relationship between incubation period and cytotoxicity in PHHs is not available. To close this
gap, 30 compounds were tested in a concentration-dependent manner for cytotoxicity in cultivated cryopreserved PHHs
(three donors per compound) for 1, 2 and 7days. The median of the EC
50
values of all compounds decreased 1.78-fold on
day 2 compared to day 1, and 1.89-fold on day 7 compared to day 1. Median values of EC
50
ratios of all compounds at day
2 and day 7 were close to one but for individual compounds the ratio increased up to almost six. Strong correlations were
obtained for EC
50
on day 1 and day 7 (R = 0.985; 95% CI 0.960–0.994), day 1 and day 2 (R = 0.964; 95% CI 0.910–0.986), as
well as day 2 and day 7 (R = 0.981; 95% CI 0.955–0.992). However, compound specific differences also occurred. Whereas,
for example, busulfan showed a relatively strong increase on day 7 compared to day 1, cytotoxicity of acetaminophen did
not increase during longer incubation periods. To validate the observed correlations, a publicly available data set, containing
data on the cytotoxicity of human hepatocytes cultivated as spheroids for incubation periods of 5 and 14days, was analyzed.
A high correlation coefficient of EC
50
values at day 5 and day 14 was obtained (R = 0.894; 95% CI 0.798–0.945). In conclu-
sion, the median cytotoxicity of the test compounds increased between 1 and 2days of incubation, with no or only a minimal
further increase until day 7. It remains to be studied whether the different results obtained for some individual compounds
after longer exposure periods would correspond better to human-repeated dose toxicity.
Keywords Hepatotoxicity· Primary human hepatocyte· Incubation period· Cell-titer-blue
Abbreviations
APAP Acetaminophen
ASP Aspirin
BOS Bosentan
BPR Buspirone
Xiaolong Gu and Wiebke Albrecht shared first authorship.
Jan G. Hengstler
hengstler@ifado.de
1
Leibniz Research Centre forWorking Environment
andHuman Factors attheTechnical University ofDortmund
(IfADo), 44139Dortmund, Germany
2
College ofVeterinary Medicine, Yunnan
Agricultural University, Kunming650201, Yunnan,
People’sRepublicofChina
3
Department ofStatistics, TU Dortmund University,
44227Dortmund, Germany
4
In Vitro Toxicology andBiomedicine, Department
ofBiology, University ofKonstanz, Universitätsstr. 10, PO
Box M657, 78457Konstanz, Germany
5
InSphero AG, Technoparkstr. 1, 8005Zürich, Switzerland
6
Unit forCell Therapy, University La Fe Hospital, Valencia,
Spain
7
Simcyp, Sheffield, UK
Konstanzer Online-Publikations-System (KOPS)
URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-2-udjshu3svswc0
Erschienen in: Archives of Toxicology ; 92 (2018), 12. - S. 3505-3515
https://dx.doi.org/10.1007/s00204-018-2302-0
3506
BUSF Busulfan
CBZ Carbamazepine
CHL Chlorpheniramine
CLON Clonidine
DFN Diclofenac
DMSO Dimethyl sulfoxide
DILI Drug-induced liver injury
EtOH Ethanol
FAM Famotidine
Glc Glucose
HYZ Hydroxyzine
INAH Isoniazid
KC Ketoconazole
LAB Labetalol
LEV Levofloxacin
MEL Melatonin
MePa Methylparaben
NAC N-Acetylcysteine
NIM Nimesulide
NFT Nitrofurantoin
PhB Phenylbutazone
PMZ Promethazine
PPL Propranolol
RIF Rifampicin
TSN Triclosan
VPA Valproic acid
Vit C Vitamin C
Introduction
Drug-induced liver injury (DILI) is one of the principal
reasons for drug withdrawal from the market (Godoy etal.
2013; Hewitt etal. 2007). DILI also belongs to the most
frequent causes of acute liver failure in industrialized coun-
tries (Bernal etal.
2010; Ostapowicz etal. 2002; Wilke etal.
2007), and occurs despite conducting standard preclinical
tests, such as subchronic and chronic rodent studies.
Currently, primary human hepatocytes represent the gold
standard model for invitro testing of drug metabolism and
cytotoxicity (LeCluyse
2001). For cytotoxicity testing, incu-
bation periods of 1 or 2days are usually used (Arbo etal.
2016; Ghallab etal. 2016); however, it has been reported
that longer incubation periods may influence the test result
(Proctor etal.
2017). A study that systematically analyzes
the relationship between incubation period and cytotoxic-
ity in primary human hepatocytes (PHHs) is not yet avail-
able. Therefore, we selected a set of 30 compounds, mostly
pharmaceutical drugs, to perform concentration and time
dependent incubations of cultivated hepatocytes. The aim of
the present study was to compare cytotoxicity after 1, 2 and
7days of test compound exposure to understand, if EC50
values change depending on the length of the incubation
period, and whether this change is by a similar factor for all
compounds or if large compound specific differences occur.
The relevance of the observed effects for toxicological rou-
tine testing will be discussed.
Materials andmethods
Chemicals andcells
Williams medium E, penicillin/streptomycin solution, Sera-
Plus (FCS), and stable
L
-glutamine were purchased from
PAN Biotech (Aidenbach, Germany), Gentamicin (10mg/
mL) was obtained from Invitrogen Corp. (Karlsruhe, Ger-
many) and insulin supplement (ITS), dexamethasone, trypan
blue solution, and all test compounds except ethanol were
purchased from Sigma-Aldrich (St. Louis, USA). Ethanol
was obtained from VWR chemicals (Mannheim, Germany),
and rat-tail tendon collagen I for monolayer culture was
obtained from Roche (Mannheim, Germany). Cell-Titer-
Blue Cell Viability assay was purchased from Promega
(Mannheim, Germany), and cryopreserved primary human
hepatocytes were obtained from BioreclamationIVT (Balti-
more, USA) (details in Supplement 1).
Cell culture
Cryopreserved primary human hepatocytes (PHH) from
three donors (AFJ, IAN and MSW) (BioreclamationIVT)
were used. Cultivation of cryopreserved hepatocytes was
performed according to a published standard operating pro-
cedure (SOP) (Godoy etal.
2013) with modifications. The
SOP used for the current study is available in Supplement 2.
Briefly, for the gel preparation, a bottle of 10mg lyophilized
rat-tail collagen was dissolved overnight in 40mL 0.2% ace-
tic acid at 4°C. Each well of 96-well plates was coated with
100μL (250μg/mL) collagen solution. The collagen solu-
tion was removed immediately, and the plates were left to
dry overnight under the cell culture hood. The plates coated
with collagen were washed three times with sterile PBS.
Cryopreserved primary human hepatocytes were thawed
in a water bath (37°C) and immediately transferred into
a Falcon tube with culture medium containing 10% FCS.
After cell counting using Trypan blue to determine viability,
50,000 cells in FCS-containing medium were plated into
each well of 96-well plates and kept at 37°C for at least
3h. For homogenous distribution of the cells, the plate was
gently shaken every 5–10min during the first half hour of
incubation. After the attachment period of 3h, the cells were
washed with warm sterile PBS three times, and 200μL FCS
free culture medium was added per well, which was kept at
37°C in the incubator overnight.
3507
Incubation ofprimary human hepatocytes withtest
compounds andcytotoxicity test
The day after plating, the cells were exposed to the con-
centrations of test compounds as indicated in the “
Results”
section. If solubility was sufficient, the concentration range
was adjusted to include at least one cytotoxic (below EC
50
)
concentration. For some compounds, the used concentra-
tions were limited by the maximal solubility of the corre-
sponding compound. In case of water soluble compounds,
substances were dissolved in medium and sterile filtered
using 0.22-μm membrane filters before adding to the cells.
For compounds not sufficiently soluble in culture medium,
DMSO was used as a solvent. If sufficient, 0.1% DMSO was
used. Only when the cytotoxic test compound concentrations
were not reached with 0.1% DMSO, the solvent concentra-
tion was increased to 0.5%. The applied DMSO stock solu-
tions for each compound and the tested compound concen-
trations are given in Suppl. Table1A and B. The DMSO
stock solutions were added to the culture medium to obtain
the final test concentrations. The culture medium contains
11mM glucose. Therefore, the indicated concentrations of
glucose were added additionally to this basal glucose con-
centration. For single exposure, the cells were incubated
with compounds for 24h or 48h; for repeated exposure,
the compound-containing medium was renewed every 48h
and the cells were incubated for a total of 7days. The cyto-
toxicity test (Cell-Titer-Blue) was performed according to
an optimized SOP (Supplement 3) and the manufacturer’s
instructions. Briefly, after the cells were incubated with
compounds for the indicated time periods, the compound-
containing medium was removed, cells were washed with
warm sterile PBS three times, and 100μL fresh FCS-free
medium with 20% Cell-Titer-Blue® reagent was added to
each well. After 3h, the supernatant was transferred to black
polystyrene 96-well plates and the fluorescence intensity
was detected with the Tecan Infinite M200 Pro plate reader
using the i-control software. Cells cultivated with culture
medium only or with the solvent DMSO only (0.1% and
0.5% DMSO; Suppl. Table1) were used as a reference for
100% viability. The applied solvent concentrations of 0.1%
or 0.5% DMSO did not cause any cytotoxicity compared to
cells cultivated in medium without DMSO. Cell viability
was calculated after background subtraction and expressed
as percentage of control. PHH of three donors were used as
three biological replicates and for each donor four technical
replicates were analyzed for the fluorescence read out.
Statistical analysis
The raw data were initially processed as follows: background
controls (fluorescence values from Cell-Titre-Blue Reagent®
mixed with medium that was not in contact with cells) were
subtracted from each data point. Replicates of control values
and of exposed samples were averaged (donor wise). Next,
averaged exposed samples were divided by the correspond-
ing averaged control values and multiplied by 100 to obtain
a percentage. Based on the assumption that the response
dependency of the concentration can be described by a sig-
moidal curve, a four-parameter log-logistic model (4pLL)
was fitted to the data. Due to the non-linearity of the 4pLL
model, the function was approximated according to the least
square method with the Gauss–Newton algorithm. Then the
individual values of each concentration were divided by the
value of the left asymptote of the fitted curve. Again, a 4pLL
model was fitted to the data. EC
20
and EC
50
values were
calculated as the concentrations where the sigmoidal curve
attains the values 80 and 50%, respectively. The advantage
of this fitting procedure is that the left asymptote is used
as a control level for calculation of EC
50
and EC
20
values
which are more robust than just using the values of the sol-
vent controls. For calculating confidence intervals of the EC
values, the concept of the ALEC (absolute lowest effective
concentration) was used (Grinberg
2017). EC
20
and EC
50
values were calculated as the concentrations where the sig-
moidal curve attains the values 80% and 50%, respectively.
Confidence intervals of the EC values were calculated using
ALEC (absolute lowest effective concentration) and the delta
method, as described in Grinberg
2017. EC values above
the highest tested concentration, as well as for cases where
the left asymptote lies above 80 or 50%, respectively, were
recorded as “>highest concentration”. The EC values of the
three donors were summarized as follows: the median was
used when all three EC values were smaller than the highest
tested concentration. When two of the three EC values were
below the highest tested concentration, the second lowest
value was used. When only one of the three EC values was
below the highest tested concentration, values of the three
donors were summarized as “>highest concentration”. The
latter values were only included as descriptive measures
into the “
Results” section but were excluded from further
calculations, such as the establishment of the box plots in
Fig.
2c and the correlation analyses in Figs.3 and 4. For the
test compound melatonin, day 1, the value of the first donor
was omitted and the mean of the two other values was used.
Pearson correlation coefficients were calculated between the
logarithmic EC summary values for two incubation times.
An upper and lower threefold deviation line was added to the
correlation plots. The upper deviation lines were obtained
by adding three (log10 scale) to the Y-axis distance of the
regression line. Correspondingly, the lower deviation line
was obtained by subtracting three (log10 scale) from the
Y-axis distance of the regression line. The statistical analy-
ses were performed with the statistical programming lan-
guage R-version 3.1.1. [
https ://www.R-proje ct.org/]. For
fitting concentration–response curves, the R-package drc
3508
(http://journ als.plos.org/ploso ne/artic le?id=10.1371/jour n
al.pone.01460 21
) was used.
Results
Selection oftest compounds andconcentration
ranges
In total, 30 compounds were selected for time- and con-
centration-dependent cytotoxicity testing (Suppl. Table1).
In addition to pharmaceuticals, ethanol, glucose, vitamin C
and DMSO were also tested. For half of the selected com-
pounds an increased risk of hepatotoxicity was previously
reported, whereas no increased risk has been reported for the
further 15 compounds at therapeutic doses (Suppl. Table1).
The tested concentrations and solvents are given in Suppl.
Table1.
Influence oftheincubation period oncytotoxicity
For cytotoxicity testing, cryopreserved hepatocytes from
three donors were used, whose characteristics are sum-
marized in Table
1. Three representative examples of the
cytotoxicity tests are given in Fig.
1, and all results and raw
data are available in the supplement (Suppl. Fig.1; Suppl.
Table2–4). The examples of propranolol (PPL), levofloxa-
cin (LEV) and clonidine (CLON) illustrate a concentration-
dependent increase in cytotoxicity for all three exposure
periods (Fig.
1a–c). In Fig.1, the EC
50
is illustrated. Simi-
larly, the EC
20
values were determined (Suppl. Fig.1).
To summarize the data, bar plots were generated for
median EC
50
values of all compounds (Fig.2a). The fig-
ure illustrates the large, more than 1000-fold, differences in
EC
50
values among the compounds, with ketoconazole and
promethazine representing the most, and dimethyl sulfox-
ide, glucose and ethanol the least cytotoxic compounds for
all tested incubation periods. Next, the ratios of the median
EC
50
values for the different incubation periods were calcu-
lated (Fig.
2b). Acetaminophen is an example of a compound
for which cytotoxicity did not increase after 7 days com-
pared to 1day of incubation (but rather showed a decrease).
In contrast, busulfan shows a relatively strong increase in
cytotoxicity after 7 days compared to 1day of incubation
(Fig.
2b). The median EC
50
ratio of all compounds was high-
est (1.89-fold) for days 1–7 (Fig.
2c). The second highest
median EC
50
ratio (1.78-fold) was obtained for days 1–2.
Median values of EC
50
ratios of all compounds at day 2 and
day 7 did not show a major difference but for many indi-
vidual compounds the ratio was different from one (Fig.
2b).
The complete set of EC
50
and EC
20
values including the
ratios is given in Suppl. Fig.2 and Suppl. Table2.
For statistical evaluation, correlation plots were created
for log
10
EC
50
values with the different incubation periods
on the x- and y-axes. The longer incubation period is shown
on the y-axis (Fig.
3). Data points indicate the experimen-
tally obtained EC
50
values for the 30 compounds, for which
the median of three donors is shown. For some of the less
toxic compounds, viability was not sufficiently reduced to
calculate an EC
50
value, because their limited solubility did
not allow for the testing of higher concentrations. Toxicity
data of compounds that did not reach EC
50
or EC
20
were
not used for further calculations, such as box plot (Fig.
2c)
or correlation analyses (Figs.
3, 4). EC
50
values after 1and
7days of incubation showed a high correlation with a cor-
relation coefficient (R) of 0.985 (with 95% confidence inter-
val [0.960, 0.994]) (Fig.
3a). A high correlation coefficient
means that the relationship between days 1 and 7 is similar
across most compounds. If all compounds would lie on or
very close to the regression line, this would mean that the
EC
50
values of longer incubation periods could be deduced
from the EC
50
values of shorter incubation periods. The
majority of compounds were within a threefold deviation
range from the regression line (plotted as a parallel line on
logarithmic scale); however, the distance of individual com-
pounds from the regression line may differ by more than
fivefold (Fig.
3a). The compounds with the highest day 1/7
ratio are busulfan, famotidine and isoniazid. Similarly good
correlations were obtained for the scatter plots of day 1 ver-
sus day 2 (R = 0.964) (Fig.
3b), as well as day 2 versus day 7
(R = 0.981) (Fig.
3c). The majority of data points were below
the diagonal line, indicating higher levels of cytotoxicity for
the longer incubation period. Similar correlation plots were
obtained for both EC
50
and EC
20
(Fig.4a–c). As observed
Table 1 Characteristics of the hepatocyte donors
Detailed donor characteristics as provided by the commercial source of human hepatocytes (Supplement 1)
CVA cerebral vascular accident
Donor no. and
abbreviation
Sex Age (year) Diagnosis Height (cm) Weight (kg) Medication Alcohol
(Y/N)
Tobacco
(Y/N)
1 (AFJ) M 54 Heart failure 177.8 75 Advil, Prilosec Y Y
2 (IAN) M 48 Head trauma 177.8 70 None Y Y
3 (MSW) M 69 CVA, stroke 179.8 80 None Y Y
3509
for EC
50
, lower values were also obtained for EC
20
, when
longer incubation periods were used. Correlation coefficients
of the EC
20
values (Fig.3) were lower compared to those
of EC
20
(Fig.4) suggesting that just beginning cytotoxicity
(EC
20
) shows larger deviations from the regression line than
half-maximal (EC
50
) cytotoxicity.
Influence oftheincubation period oncytotoxicity
inpublished data
Recently, Proctor etal. (2017) published cytotoxicity data
with different incubation periods. To our knowledge, this
is the only available public data set of human hepatocytes
where cytotoxicity of a larger number of compounds
has been studied with different exposure periods (data
in supplement TableS4 of Proctor etal.
2017). Proctor
etal. incubated spheroids of human hepatocytes for 4
and 14days. However, they did not analyze correlations
between the time points in their study. A correlation analy-
sis was performed in the present study and is illustrated
in Fig.
5. The result shows a highly significant correla-
tion with a correlation coefficient of R = 0.894, which is
similar to the correlation coefficients obtained for our
data. The largest difference between day 14 and day 5 was
obtained for atorvastatin (EC
50
day 14: 0.006mM; EC
50
day 5: 0.0707mM). Surprisingly, for some compounds
lower EC
50
values were obtained for the shorter incubation
period (e.g. flutamide: EC
50
day 14: 0.0257mM; EC
50
day
5: 0.0125mM; and phenformin: EC
50
day 14: 0.0125mM;
EC
50
days 5–6: 0.0077mM).
Fig. 1 Representative examples of Cell-Titer-Blue cytotoxicity data
for propranolol (a), levofloxacin (b), and clonidine (c) in primary
human hepatocytes after incubation for 1, 2 and 7days. The corre-
sponding data for all 30 compounds are available in Suppl. Fig.1 and
Suppl. Table2–4. The concentration-dependent curves represent data
from three donors with four technical replicates each. The cell viabil-
ity for each concentration is presented as the percentage of untreated
controls. Gray symbols indicate the viability values for each technical
replicate normalized to untreated controls; whereas, black symbols
represent the mean values of all technical replicates for each concen-
tration. The vertical line indicates the concentration which causes
50% loss of viability (EC
50
). The dashed vertical line shows the 95%
confidence interval for the concentration. The resulting EC
50
values
are given for each panel. (Color figure online)
