You are on page 1of 8

Modeling microbial ethanol production by E.

coli under aerobic/
anaerobic conditions: Applicability to real postmortem cases and to
postmortem blood derived microbial cultures
Vassiliki A. Boumba
a,
*, Nikolaos Kourkoumelis
b
, Panagiota Gousia
c
, Vangelis Economou
c
,
Chrissanthy Papadopoulou
c
, Theodore Vougiouklakis
a
a
Department of Forensic Medicine & Toxicology, Medical School, University of Ioannina, 45110 Ioaninna, Greece
b
Department of Medical Physics, Medical School, University of Ioannina, 45110 Ioaninna, Greece
c
Department of Microbiology, Medical School, University of Ioannina, 45110 Ioaninna, Greece
1. Introduction
Determination of blood alcohol concentration (BAC) in post-
mortem cases is part of the death investigation process since
ethanol might be a causal or contributory factor to the manner of
death. The accurate scientific interpretation of the ethanol analysis
results presents a critical task in cases where microbial ethanol
neo-formation is suspected, or, alternatively, the origin of
postmortem ethanol (ante mortem ingestion or microbial produc-
tion) is questioned.
In order to achieve a feasible accuracy in interpreting the results
of postmortem ethanol analysis different approaches have been
suggested. The analysis of multiple specimens and from multiple
sampling sites could reveal atypical distribution of ethanol
throughout the different compartments of the dead body and
thus could be indicative of postmortem ethanol neo-formation [1–
5]. The determination of various low molecular weight volatiles in
postmortem specimens is suggested as a criterion to specify the
origin of the measured postmortem ethanol and, if certain volatiles
are detected then postmortem ethanol production should be
suspected [3,4,6–10]. Finally, the determination of various non-
oxidative metabolites of ethanol is used to distinguish the ante
mortem ingestion from the postmortem synthesis of ethanol
[4,11–15]. Common feature in the aforementioned approaches is
that the documentation, or not, of the microbially produced
ethanol arises in qualitative terms.
Recently we reported the formulation of mathematical models
for calculating the microbial neo-formed ethanol in strict
anaerobic cultures of the bacterial strains Escherichia coli,
Clostridium perfringens, and Clostridium sporogenes [16]. This
approach was the first approximation to the quantification of
the microbial ethanol production in cases where other low
molecular weight alcohols were produced simultaneously with
ethanol.
Forensic Science International 232 (2013) 191–198
A R T I C L E I N F O
Article history:
Received 26 April 2013
Received in revised form 19 July 2013
Accepted 26 July 2013
Available online 7 August 2013
Keywords:
Postmortem ethanol
Postmortem blood
Multiple linear regression
Autopsy
E. coli
Microbial ethanol
A B S T R A C T
The mathematical modeling of the microbial ethanol production under strict anaerobic experimental
conditions for some bacterial species has been proposed by our research group as the first approximation
to the quantification of the microbial ethanol production in cases where other alcohols were produced
simultaneously with ethanol.
The present study aims to: (i) study the microbial ethanol production by Escherichia coli under
controlled aerobic/anaerobic conditions; (ii) model the correlation between the microbial produced
ethanol and the other higher alcohols; and (iii) test their applicability in: (a) real postmortemcases that
had positive BACs (>0.10 g/L) and co-detection of higher alcohols and 1-butanol during the original
ethanol analysis and (b) postmortem blood derived microbial cultures under aerobic/anaerobic
controlled experimental conditions.
The statistical evaluation of the results revealed that the formulated models were presumably
correlated to 1-propanol and 1-butanol which were recognized as the most significant descriptors of the
modeling process. The significance of 1-propanol and 1-butanol as descriptors was so powerful that they
could be used as the only independent variables to create a simple and satisfactory model. The current
models showed a potential for application to estimate microbial ethanol – within an acceptable standard
error – in various tested cases where ethanol and other alcohols have been produced from different
microbes.
ß 2013 Elsevier Ireland Ltd. All rights reserved.
* Corresponding author. Tel.: +30 26510 07724; fax: +30 26510 07857.
E-mail addresses: vboumba@cc.uoi.gr, metsbou@yahoo.gr (V.A. Boumba).
Contents lists available at ScienceDirect
Forensic Science International
j our nal h o mepage: w ww. el sevi er . co m/ l oc at e/ f o r sc i i nt
0379-0738/$ – see front matter ß 2013 Elsevier Ireland Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.forsciint.2013.07.021
The present contribution describes the development of simple,
albeit accurate, mathematical models which could quantify the
microbial produced ethanol in correlation with other alcohols in E.
coli cultures under initially aerobic and then anaerobic, controlled
laboratory conditions. The applicability of the constructed models
was tested in real postmortem cases that had positive BACs
(BAC > 0.10 g/L) and co-detection of higher alcohols and 1-butanol
during the original ethanol analysis. The applicability of the
models was also tested to microbial cultures of blood culture
media inoculated with postmortem blood from autopsy cases
under aerobic/anaerobic controlled experimental conditions.
2. Materials and methods
2.1. Microbial cultures
2.1.1. Bacterial strain
The bacterial strain used in this study was E. coli ATCC 35218. The E. coli strain
was stored at À80 8C (Mikrobank beads – Prolab Diagnostics, Canada) and prior to
use it was revived by transferring 3–5 beads into 10 mL of Brain Heart Infusion (BHI)
(CM0225, Oxoid) and was incubated at 37 8C for 18–24 h. The bacterial inoculum
was adjusted to a working concentration of 10
6
cells/mL using spectrophotometry.
In brief, a standard solution of 10
8
cells/mL was prepared from an initial culture
after overnight incubation at 37 8C aerobically and the concentration was
determined using the spectrophotometer (Model 6305, Jenway, England) at
600 nm. In order to obtain a final concentration of 10
6
cells/mL, 100 mL of the
standard inoculum were added in 9.9 mL of BHI in sterile tubes. The tubes were
incubated aerobically for the first 5 days. After this period 1 mL of sterile parafine oil
was added in each tube as an overlay and the tubes were placed into GENbag anaer
incubation systems (45534, bioMe´ rieux, France) so as to provide anaerobic
conditions from day 6 to day 30. The incubation, both anaerobic and aerobic, was
performed at 25 8C for 30 days.
2.1.2. Bacterial enumeration
The bacterial concentration was determined by the pour plate count technique
[16]. Decimal dilutions were performed in tubes with 9 mL of Maximal Recovery
Diluent (MRD, 02-510, Scharlau Chemie, Spain). From each tube an inoculum of
1 mL was transferred into 2 petri dishes and 15 mL of Plate Count Agar (1.05463,
Merck KGaA, Germany) melted at approximately 45 8C were added. After
solidification all plates were incubated aerobically at 37 8C for 48 h. Colonies were
counted under a colony counter (Stuart SC6, Barloworld Scientific Ltd., UK).
2.2. Postmortem blood derived microbial cultures
2.2.1. Identification of microbes
Seven postmortem blood samples were examined for the presence of E. coli,
Staphylococcus aureus, Enterococcus spp., and Clostridium spp. Ten mL of aseptically
collected blood was added into 1 mL of Buffered Peptone Water – BPW (CM1049,
Oxoid, UK). E. coli was isolated by plating a loopful of BPW on Violet Red Bile agar
(4021852, Biolife, Italy) and Chromocult Coliform Agar (1.10426.0500, Merck KGaA,
Germany). The inoculated media were incubated aerobically at 37 8C for 24 h.
Characteristic colonies were selected, pure cultured and identified to the genus
level using the API20E system (bioMe´ rieux, France). For the isolation of Gram
positive strains (S. aureus and Enterococcus spp.), Manitol Salt agar (CM0085, Oxoid,
UK) and D-Coccosel Bile Esculin Agar (51025, bioMe´ rieux, France) were used,
respectively. The plates were incubated at 37 8C aerobically. Characteristic colonies
were selected, pure cultured and identified to the genus level by biochemical tests.
For Staphylococci the catalase test, the coagulase test (Staphytect Plus, Oxoid, UK)
and the API STAPH system (bioMe´ rieux, France) were used. Biochemical
identification of the Enterococci was performed using the catalase test and the
API 20 STREP system (bioMe´ rieux, France). For the isolation of Clostridium species a
loopful was streaked onto blood agar plates, the inoculated plates were incubated
anaerobically using the GENbag anaer incubation systems (bioMe´ rieux, France) for
48 h at 37 8C and were examined for b-haemolysis. Identification was performed
microscopically after Gram stain and with the use of the API 20A system
(bioMe´ rieux, France).
2.2.2. Normal human blood culture media inoculated with postmortem blood
(postmortem blood derived microbial cultures)
The postmortem blood derived cultures were performed by a procedure based to
that described by Appenzeller and colleagues [15] modified to fit the conditions of
microbial cultures used herein. Specifically, ethanol free whole human blood from
healthy individuals was inoculated with postmortem blood (100 mL/mL of normal
human blood). Postmortem blood samples were collected into sterilized collection
tubes containing anticoagulant, during autopsy from cases with putrefaction and
also were subjected to the procedure of microbial identification (described in
Section 2.2.1). The blood samples were kept at 4 8C until further use. One mL
portions of the inoculated blood were transferred into 40 sterilized blood tubes
(Vacuette EDTAK3, 3 mL). Twenty tubes were spiked with the appropriate volume
of a concentrated glucose solution (sterile dextrose 5% (w/v), pyrogen free solution
for intravenous infusion, DEMO Ltd., Greece) so as to obtain 200 mg/dL final glucose
concentration. The other tubes were incubated without additional glucose. All the
tubes were capped and incubated at 25 8C. On the seventh day a sterile parafin oil
overlay was added into each tube and the tubes were placed into the GENbag anaer
incubation system. At days 0, 1, 2, 3, 5, 7, 9, 11, 15 and 20 two tubes were removed
from the incubation system and alcohols concentrations were determined by HS-
GC–FID.
2.3. Determination of volatiles
The alcohols concentration in the culture medium and the blood samples were
determined by the previously described procedure [16].
2.4. Statistical analysis
Regression analysis was employed to model the correlation between the
microbial produced ethanol and the other higher alcohols. The experimental data
were consistent with a Gaussian distribution (normality test) and were analyzed
with multiple linear regression (MLR) methods to model the relationship between
ethanol concentrations as the dependent variable being in linear correlation with
the concentrations of the other alcohols (independent variables). Each descriptor’s
relevance to the current model was assessed by partial least squares (PLS)
regression utilizing the PLS1 algorithm. During the analysis, one outlier value was
removed from the dataset.
3. Results and discussion
3.1. Experimental determination of alcohols produced in pure E. coli
cultures
E. coli is a Gram negative, aerobic and facultative anaerobic, rod-
shaped bacterium that is commonly found in the lower intestine of
warm-blooded organisms. It is considered one of the main
colonizers in corpses and of the main ethanol producers [17].
The growth curve of E. coli during the 30 days incubation at 25 8C is
presented in Fig. 1. During the first 5 days of incubation the
conditions were aerobic. On the fifth day up to the end of the
incubation the conditions of microbial growth changed to
anaerobic. By the fifth day of incubation the bacterial population
reached the stationary phase of growth. The change from aerobic to
anaerobic conditions was selected in order to approximate the
usual conditions where a postmortem blood evolves from death to
autopsy (initially air is available, which gradually decreases and
finally depletes) [1,19] and the duration of changing was selected
in order to obtain the optimal microbial growth. At 4 8C no growth
was observed as expected (not shown) [15,16,18].
The amounts of ethanol, and other alcohols produced under the
applied experimental conditions were variable (Table 1). Ethanol
1,0E+06
1,0E+07
1,0E+08
1,0E+09
1,0E+10
30 25 20 15 10 5 0
Days
B
a
c
t
e
r
i
a
l

c
o
u
n
t

(
L
o
g

1
0

C
F
U

/

m
L
Fig. 1. Growth curve of E. coli showing the number of bacteria during the incubation
period of 30 days at 25 8C. The time of change between aerobic and anaerobic
incubation is shown by the dotted line at 5 days.
V.A. Boumba et al. / Forensic Science International 232 (2013) 191–198 192
was produced in higher concentrations compared to the other
alcohols. Considering the ethanol concentration at days 0 and 5
(0 g/L and 0.55 g/L, respectively), it is concluded that the mean rate
of ethanol increase is 0.11 g/L/day (or 11 mg/dL/day) during the
first five-days period of incubation. After switching to anaerobic
conditions, from day six to day nine, the rate of ethanol production
decreased to a mean rate of 0.017 g/L/day (or 1.7 mg/dL/day).
Afterwards, from day 10 to the end of the experiment, ethanol kept
a plateau concentration of about 0.62 g/L.
The most abundantly produced alcohol (except ethanol) was 1-
propanol which was continuously increasing during the first 17
days. Interestingly, the mean rate of 1-propanol’s increase was
0.42 mg/dL/day during the first five days, while for the next twelve
days the increase rate declined, and afterwards, from day 18 to the
end, 1-propanol showed no considerable change. The third in
abundance alcohol was 1-butanol which increased the first five
days with a rate approximately 10 times slower (0.042 mg/dL)
than the respective increase rate of 1-propanol, while afterwards it
remained practically stable. Finally, isobutanol and methyl-
butanol increased slowly for twelve days and then remained
stable.
Our results show that 1-propanol and 1-butanol are the main
alcohols that are produced simultaneously with ethanol by
microbes under the applied experimental conditions. The amounts
of ethanol and other alcohols produced by E. coli in BHI cultures
under combined aerobical/anaerobical conditions were higher
compared to those produced by E. coli in BHI cultures under strict
anaerobic conditions [16]. The observed discrepancies are
attributed to the presence/absence of air. Therefore it is observed
that higher amounts of alcohols are produced by certain microbes
(here E. coli) under aerobic conditions.
It is of importance to underline that 1-butanol was produced
during the first five days of incubation, in the presence of air, while
during this period both ethanol and 1-propanol had the higher
production rates; ethanol and 1-propanol continued to accumu-
late, with a decreased production rate, for further 4 and 12 days,
respectively, under anaerobic conditions. To explain these results
the following aspects should be considered: (i) ethanol can be
produced from all the available postmortem substrates but with
higher rates and yields from carbohydrates; (ii) 1-propanol and 1-
butanol (and the other higher alcohols) could be produced from all
the available postmortem substrates through branched biochemi-
cal pathways that depend on glucose metabolism; and (iii) 1-
propanol, isobutanol and methyl-butanols are produced using
amino acids as substrates [20]. Therefore, it should be concluded
that carbohydrates (as the preferable substrate) were consumed by
the microbes during the aerobic phase of growth, generating
mainly ethanol, 1-propanol and 1-butanol. As carbohydrates were
being depleted from the medium, other available substrates were
consumed continuing to generate higher alcohols (1-propanol,
isobutanol and methyl-alcohols) and ethanol in line with previous
suggestion [20].
3.2. Formulation of models for ethanol production by E. coli
3.2.1. Statistical evaluation of volatiles
All volatiles during the first 5 days follow a linear approxima-
tion (r
2
> 0.9, not shown here) with the exception of ethanol. For
the next days (6–30), MLR is not able to successfully fit a similar
linear model (smaller r
2
value) regarding all variables with the
notable exception of 1-propanol. Probably, 1-propanol plays a
significant role in the successful model prediction while all data is
better to be treated simultaneously since for the first 5 days, data is
ambiguous regarding the variance. By performing a principal
component analysis (PCA) for the whole data to determine how
original variables are projected to a smaller set of underlying
variables, it becomes evident that ethanol explains 99% of the total
variance while the remaining 1% is explained by 1-propanol.
Table 1
Alcohols concentration produced during the 30 days incubation period of E. coli in BHI culture medium at 25 8C under aerobic (days 1–5) and anaerobic (days 6–30)
conditions.
Day Ethanol (g/L) 1-Propanol (mg/dL) Isobutanol (mg/dL) 1-Butanol (mg/dL) Methyl-butanol (mg/dL)
0 0 0 0 0 0
1 0.49 1.12 0.01 0.1 0.03
2 0.48 1.33 0.01 0.15 0.03
3 0.50 1.50 0.01 0.18 0.04
4 0.52 1.81 0.01 0.19 0.04
5 0.55 2.08 0.02 0.21 0.05
6 0.53 2.05 0.02 0.21 0.05
7 0.56 2.18 0.02 0.22 0.06
8 0.59 2.30 0.02 0.23 0.04
9 0.61 2.33 0.02 0.23 0.05
10 0.62 2.85 0.02 0.24 0.06
11 0.58 2.67 0.02 0.23 0.06
12 0.60 2.87 0.03 0.24 0.08
13 0.63 3.18 0.03 0.24 0.08
14 0.60 2.63 0.03 0.23 0.08
15 0.58 2.78 0.02 0.23 0.07
16 0.61 3.04 0.03 0.23 0.11
17 0.61 3.22 0.03 0.23 0.07
18 0.62 3.01 0.03 0.23 0.07
19 0.62 3.15 0.03 0.23 0.07
20 0.63 3.47 0.03 0.24 0.08
21 0.62 3.23 0.03 0.24 0.08
22 0.62 3.30 0.03 0.22 0.06
23 0.61 3.18 0.03 0.24 0.07
24 0.61 3.07 0.03 0.24 0.08
25 0.61 3.08 0.03 0.23 0.07
26 0.61 3.16 0.03 0.23 0.08
27 0.61 3.56 0.03 0.23 0.09
28 0.60 3.13 0.03 0.23 0.07
29 0.61 3.20 0.03 0.24 0.08
30 0.62 3.63 0.03 0.24 0.09
V.A. Boumba et al. / Forensic Science International 232 (2013) 191–198 193
Therefore, it is shown again that the most important variations
arise from ethanol and 1-propanol. Finally, by subtracting one by
one variable, it is concluded that isobutanol and methyl-butanol
are the variables that do not significantly contribute to the total
variation.
3.2.2. Modeling whole data
By performing a PLS model the descriptors relevance becomes
also evident as it is shown in Fig. 2A. The significance of each
alcohol in building the mathematical models is in relevance with
its produced amount.
The corresponding model (r
2
= 0.90) is described by Eq. (1A):
Ethanol ¼ 0:05 Â 1-propanol þ 1:13 Â isobutanol þ 0:95
 1-butanol À 1:60  methyl-butanol þ 0:31 (1A)
By modeling the data with only the most significant variables
(1-propanol and 1-butanol) the prediction shows a considerably
satisfactory linear correlation (r
2
= 0.75), indicating the power of
these two alcohols as descriptors of the process, as it is shown in
Fig. 2B. Therefore we obtain the following model:
Ethanol ¼ 0:05 Â 1-propanol þ 0:53 Â 1-butanol þ 0:32 (1B)
3.2.3. Modeling data from day 1 to 5
If we analyze the data separately for the first 5 days of aerobical
incubation, we see that the ethanol concentration is heavily
dependent on 1-propanol, followed by isobutanol, and then by 1-
butanol concentration (Fig. 3). The variance of methyl-butanol is of
least significance. The relevant PLS model is described by the Eq. (2):
Ethanol ¼ 0:07 Â 1-propanol þ 3:77 Â isobutanol À 0:26
 1-butanol þ 0:07  methyl-butanol þ 0:39 (2)
3.2.4. Modeling data from day 6 to 30
Using similar methodology and the data for the anaerobic
cultivation period from day 6 to day 30 the PLS model is as
described by Eq. (3):
Ethanol ¼ 0:05 Â 1-propanol þ 1:06 Â isobutanol þ 1:76
 1-butanol À 1:62  methyl-butanol þ 0:14 (3)
The descriptors relevance shows again the significance of 1-
propanol and 1-butanol in describing the process which is very
similar with the whole dataset (Fig. 4).
3.2.5. Validation modeling with PLS
Validation modeling was performed by creating a small dataset
of 13 randomly selected values of measurements made every 3
days (validation dataset). The rest of the data values were used as
the training dataset. The resulted PLS model is described by the
following equation:
Ethanol ¼ 0:06 Â 1-propanol þ 1:34 Â isobutanol þ 1:00
 1-butanol À 1:74  methyl-butanol þ 0:29
This equation was applied to the validation dataset with
r
2
= 0.93 showing that the model for the whole dataset can indeed
successfully describe the ethanol production from E. coli under the
applied experimental conditions confirming our results.
3.2.6. Aspects of microbial ethanol production modeling
The models for estimating ethanol production by E. coli under a
combination of aerobic/anaerobic conditions were presumably
correlated to 1-propanol and 1-butanol, as the relevant descrip-
tor’s significance revealed. This result is in discordance with the
models constructed previously for E. coli under strict anaerobic
conditions, where the main descriptors of ethanol modeling were
1-butanol followed by methyl-butanol. On the other hand, the
result is in concordance with the models constructed previously
for C. perfringens and C. sporogenes [16].
It is worth mentioning that 1-propanol [6,8–10] and 1-butanol
[7] have been recognized, in previous studies by other authors to
be correlated with the postmortem ethanol production basically in
qualitative terms. Our results confirm and enhance the significance
of the presence of 1-propanol and 1-butanol as main indicators of
microbial ethanol production and further, recognize them as the
main descriptors of the quantification process. The relevance of
both these alcohols as descriptors of ethanol modeling is so
powerful that the model constructed by using only these two
variants is quite satisfactory and in agreement with the general
consideration that the simplest the model the easier its applica-
tion.
3.3. Applicability of the models to death-related forensic cases
3.3.1. Real postmortem cases
The applicability of the models presented in this report was
tested in 60 postmortem cases. The cases were selected by
reviewing retrospectively our chromatogram archives and select-
Fig. 2. (A) Four descriptors (4D) model and (B) two descriptors (2D) model for ethanol production by E. coli by analyzing the whole dataset for the 30 days of incubation.
V.A. Boumba et al. / Forensic Science International 232 (2013) 191–198 194
ing those that had positive blood alcohol concentrations
(BAC 0.10 g/L) and co-detection of higher alcohols and 1-butanol
during the original ethanol analysis. The concentrations of 1-
butanol and higher alcohols were calculated for the selected cases.
Cases were classified according to the manner of death – as
resulted from the retrospective review of the forensic pathologist’s
report – as natural (11 cases, 18.3%), violent (14 cases, 23.3%) or
undetermined (35 cases, 58.3%) cases. Also, it was revealed that 58
out of the 60 samples were from corpses with putrefaction, one
from a corpse with extended trauma and one from a totally
carbonized body. In Table 2 are presented the measured BACs, the
estimated ethanol concentrations after applying each model
(Eqs. (1A), (1B), (2) and (3)) and the relative standard error for
each one of the selected cases. The higher acceptable standard
error (in order to consider successful the application of the models)
in the estimation of the produced ethanol when compared to the
measured original BACs, was selected to be 40%, the same as in our
previous relevant study [16].
The model described by Eq. (1A) estimated successfully the
microbial produced ethanol in 24 out of the 60 cases (38%), the
model described by Eq. (2) estimated successfully the microbial
produced ethanol in 16 out of the 60 cases (25%), and the model
described by Eq. (3) estimated successfully the microbial produced
ethanol in 22 out of the 60 cases (35%). However, the simplest
model, described by Eq. (1B) that was constructed by only two
descriptors 1-propanol and 1-butanol was applied successfully in
the majority of cases (26 out of 60 cases, 42%).
The overall picture was that at least one of the current models
(Eqs. (1A), (1B), (2) and (3)) were applied successfully in: 41 out of
the 60 cases (68%), 39 of them with putrefaction (39/41, 95%); in
28/43 cases (65%) with BAC lower than 0.7 g/L; and, in 13 out of 17
cases (76%) with BAC higher than 0.7 g/L. These results are
consistent with the general expressed consideration that decom-
position is an aggravating factor for postmortem ethanol produc-
tion and, in line with the statement that concentrations of ethanol
lower than 0.7 g/L are most probably due to microbial ethanol neo-
formation [2,21,22].
3.3.2. Postmortem blood-derived microbial cultures
Post-mortem blood samples from seven autopsy cases were
used to inoculate normal human blood, with and without
additional glucose, as described in Section 2.2. The initial glucose
Fig. 3. Four descriptors (4D) model for ethanol production by E. coli by analyzing the data for the first five days of incubation under aerobic conditions.
Fig. 4. Four descriptors (4D) model for ethanol production by E. coli by analyzing the data from day 6 to day 30 of incubation under anaerobic conditions.
V.A. Boumba et al. / Forensic Science International 232 (2013) 191–198 195
concentration of the spiked blood samples was chosen to be
200 mg/dL in order to be the same with the glucose content of the
BHI culture medium. The selection of cases was randomized
amongst routinely autopsied corpses with obvious signs of
putrefaction at autopsy. The characteristics of the selected cases
(microbes detected, original BACs, estimated BACs and relative
standard error ranges after applying the models to the autopsy
samples) and, the results after applying the models described by
Eqs. (1B) and (2) to the inoculated samples for each case (mean
estimated BACs and standard deviations, relative standard error
ranges) are presented in Table 3. Five out of the seven blood
samples carried microbial strains (E. coli, Enterococcus faecalis,
Klebsiella species) commonly found in human corpses [1,23].
Clostridium species were not identified in blood samples from cases
1, 5 and 6, while for the cases 2, 3, 4 and 7 the results were
inconclusive. Four out of the seven blood samples were positive for
ethanol and other alcohols during the original ethanol analysis
(these were presented as cases 15, 37, 50 and 59 in Table 2). When
Table 2
Application of each model constructed for E. coli (Eqs. (1A), (1B), (2) and (3)) in postmortem cases to calculate the microbially produced ethanol. The original ethanol
concentration measured for each case along with the manner of death are also provided.
Case no. Manner of death Measured ethanol (g/L) Estimated ethanol (g/L) Relative standard error (E%)
Eq. (1A) Eq. (1B) Eq. (2) Eq. (3) Eq. (1A) Eq. (1B) Eq. (2) Eq. (3)
1 Undetermined 0.10 0.35 0.37 0.45 0.17 À244 À259 À337 À68
2 Undetermined 0.10 0.61 0.63 0.77 0.41 À489 À504 À637 À294
3 Natural 0.16 0.59 0.45 1.01 0.25 À268 À184 À530 À58
4 Natural 0.16 0.35 0.36 0.44 0.17 À113 À123 À171 À3
5 Violent 0.19 0.44 0.45 0.58 0.26 À135 À138 À208 À36
6 Natural 0.20 0.33 0.34 0.42 0.15 À64 À70 À111 25
7 Violent 0.20 0.60 0.60 0.77 0.39 À198 À201 À283 À97
8 Natural 0.21 0.51 0.50 0.71 0.32 À145 À138 À236 À52
9 Undetermined 0.21 0.54 0.41 1.00 0.35 À158 À94 À376 À66
10 Natural 0.22 0.47 0.48 0.60 0.28 À116 À121 À178 À29
11 Undetermined 0.25 0.38 0.40 0.48 0.20 À52 À57 À92 21
12 Undetermined 0.27 0.31 0.32 0.39 0.13 À13 À19 À44 52
13 Violent 0.28 0.68 0.67 0.79 0.53 À148 À142 À187 À92
14 Undetermined 0.29 0.49 0.49 0.63 0.30 À70 À72 À121 À3
15 Violent 0.33 1.00 0.81 0.55 1.20 À200 À141 À63 À261
16 Undetermined 0.33 1.34 0.97 0.72 1.76 À304 À193 À117 À433
17 Violent 0.34 0.66 0.62 0.64 0.57 À92 À81 À86 À66
18 Undetermined 0.35 0.62 0.52 1.03 0.42 À78 À49 À194 À20
19 Natural 0.38 0.31 0.32 0.42 0.14 17 15 À10 63
20 Natural 0.39 0.77 0.56 1.34 0.23 À98 À43 À245 42
21 Undetermined 0.39 0.40 0.42 0.51 0.22 À2 À6 À29 45
22 Undetermined 0.42 0.84 0.85 1.05 0.61 À101 À104 À151 À47
23 Violent 0.42 0.67 0.66 0.88 0.46 À59 À57 À110 À9
24 Natural 0.42 0.50 0.52 0.63 0.31 À20 À23 À51 26
25 Natural 0.42 0.38 0.58 0.70 0.52 9 À38 À67 À23
26 Undetermined 0.47 1.11 0.81 2.03 0.87 À137 À73 À333 À85
27 Violent 0.52 0.74 0.74 0.95 0.52 À43 À43 À84 À1
28 Undetermined 0.54 0.31 0.32 0.39 0.13 44 41 28 76
29 Undetermined 0.58 1.36 0.76 2.97 1.10 À135 À32 À414 À90
30 Undetermined 0.59 0.49 0.49 0.64 0.29 18 17 À8 50
31 Undetermined 0.60 0.76 0.71 1.07 0.54 À27 À20 À79 9
32 Undetermined 0.60 0.36 0.38 0.52 0.18 40 37 14 70
33 Undetermined 0.61 0.79 0.81 0.98 0.57 À29 À32 À61 7
34 Violent 0.62 1.09 0.96 1.65 0.84 À76 À55 À166 À36
35 Undetermined 0.62 0.66 0.68 0.83 0.46 À7 À9 À34 27
36 Undetermined 0.63 1.19 0.93 2.06 0.94 À89 À48 À226 À49
37 Undetermined 0.64 0.63 0.61 0.72 0.48 2 5 À13 26
38 Violent 0.65 1.02 0.96 1.03 0.95 À57 À48 À58 À45
39 Violent 0.66 0.51 0.40 0.91 0.32 22 39 À38 51
40 Undetermined 0.66 0.28 0.57 2.15 1.35 57 14 À224 À104
41 Undetermined 0.68 0.93 0.78 1.20 À0.27 À36 À14 À76 139
42 Undetermined 0.69 0.48 0.46 0.68 0.29 30 33 1 58
43 Violent 0.69 0.79 0.54 1.55 0.57 À12 24 À119 19
44 Undetermined 0.71 0.66 0.57 1.04 0.45 7 19 À46 36
45 Undetermined 0.71 0.48 0.49 0.62 0.29 31 32 12 59
46 Natural 0.72 0.75 0.75 1.11 0.09 À4 À4 À54 88
47 Undetermined 0.73 0.79 0.74 1.10 0.57 À7 À2 À50 22
48 Undetermined 0.75 0.96 0.98 1.19 0.73 À28 À30 À59 3
49 Undetermined 0.81 0.63 0.59 1.02 0.12 23 28 À25 85
50 Violent 0.84 0.88 0.87 1.15 0.65 À5 À4 À37 21
51 Undetermined 0.87 1.23 1.11 1.76 0.25 À41 À28 À102 72
52 Undetermined 0.91 0.91 0.92 1.14 0.68 À1 À2 À26 25
53 Undetermined 0.94 2.13 0.82 5.42 1.81 À127 13 À477 À93
54 Undetermined 0.98 0.73 0.56 1.30 0.52 26 43 À32 47
55 Violent 1.06 0.42 0.44 0.55 0.25 60 59 48 77
56 Undetermined 1.08 0.49 0.49 0.65 0.30 55 55 40 72
57 Violent 2.15 0.98 0.95 1.05 0.85 55 56 51 61
58 Undetermined 2.58 1.10 1.00 1.59 0.85 57 61 38 67
59 Undetermined 2.94 0.83 0.80 0.97 0.70 72 73 67 76
60 Natural 3.56 0.78 0.78 1.01 0.56 78 78 72 84
V.A. Boumba et al. / Forensic Science International 232 (2013) 191–198 196
ethanol and other alcohols were produced during the incubation of
the inoculated samples the procedure was considered positive and
the applicability of the models was tested. Therefore the procedure
was positive only for two out of the seven tested postmortem
bloods (cases 1 and 2 from Table 3).
The application of the current models for the original autopsy
blood of the first case, where E. coli was identified, gave acceptable
standard errors (selected to be <40%). In the inoculated blood
samples in the presence of additional glucose higher amounts of
ethanol were produced and, the models gave better scores when
applied to these samples compared to the inoculated blood
samples without additional glucose (Table 3). These results show
that the current models could be used to estimate neo-formed
ethanol by E. coli either under unspecified environmental
conditions or in laboratory conditions. The observed discrepancies
should be apparently attributed to the differences in the glucose
content of the medium and possibly to differences in the
succession of microbes in the inoculated blood samples compared
to the original blood.
In the second case the models estimated that only a portion of
the detected ethanol could be of microbial origin. Since E. faecalis is
not an ethanol producing microorganism [16] we should assume
that another microbe (not identified) was activated in the original
blood, from death to autopsy, and then during incubation of the
inoculated samples. In the inoculated blood samples with
additional glucose, more ethanol was produced compared to the
samples without additional glucose. Interestingly, in this latter
case, the application of the models succeeded better scores than
those obtained for the inoculated samples with glucose. These
results show that the current models could be used successfully to
estimate the neo-formed ethanol in cases where ‘‘ethanol
producing’’ microbe(s) (different than E. coli) had grown either
under unspecified environmental conditions or under laboratory
conditions.
In the autopsy blood samples from cases 3 and 4, Klebsiella
species had produced the alcohols to each autopsy blood sample.
Previously, it was reported that Klebsiella pneumoniae was capable
of producing ethanol in urine spiked with glucose, by fermentation
at 35 8C [24]. However, the application of the current models was
within acceptable standard error range only in case 4 (Table 3). The
results suggest that in cases where Klebsiella species were
activated, occasionally, the models could be used effectively.
Alternatively, we cannot exclude the possibility that more
microbe(s) could have grown to each original blood, and had
produced the alcohols. Finally, the inoculation procedure was
negative for cases 3 and 4, probably indicating that the ‘‘ethanol
producing’’ microbes that generated the alcohols in the original
blood (Klebsiella species or other) could not been regenerated to
the inoculated samples under the applied experimental conditions.
Postmortem blood samples that were negative for ethanol in the
original volatile analysis (cases 5–7) did not produce alcohols during
incubation of the inoculated samples, irrespectively whether
microbes had been detected (case 5) or not (cases 6 and 7). For
these last three cases it can be concluded that there were not present
‘‘ethanol producing’’ microbes in the autopsy bloods and, conse-
quently, there was not alcohols production during the incubation
period of the postmortem blood derived cultures.
In general, these results indicate that whenever a blood sample
is contaminated with microbe species that are ‘‘ethanol produ-
cers’’, a profile of ethanol and other alcohols would be generated if
the conditions are favorable for microbial growth. In such a case
the current models show a potential for application, irrespectively
of the activated microbe(s). Influencing factors appear to be the
culture composition and more specifically the glucose content, the
conditions of microbial growth and probably, the succession of
microbes.
3.3.3. Aspects, limitations and outcomes of the application of models
The working hypothesis for our study is that although the
postmortem or putrefactive conditions could be extremely
different and variable among different cases, the patterns of other
alcohols would follow more or less, in qualitative and quantitative
terms, the ethanol production, since the biochemical pathways of
their microbial production are interactive [20]. The patterns of
produced ethanol and other alcohols, under controlled experi-
mental conditions, could be described by mathematical equations
(models); the patterns of ethanol and alcohols produced under
different, real conditions, could be approximated by the models,
and this is why ethanol estimation should be made within an
acceptable standard error.
Ideally, a model should exist for every postmortem case
(predisposing that it was constructed under the same conditions)
which is impossible, since the conditions could not be defined
accurately. In our view, each model represents a ‘‘tool’’ which could
be used as an alternate choice to estimate the microbially produced
ethanol in real cases. In every given case, the effectiveness of each
Table 3
Characteristics of the autopsy bloods used to inoculate normal human blood (original BACs, estimated BACs and relative standard error ranges (jE%j)) and results for the
inoculated samples with and without additional glucose in each case (produced ethanol concentration, relative standard error ranges after applying the models described by
Eqs. (1B) and (2), respectively).
Case Microbes detected BAC (g/L) Estimated
BAC range (g/L)
Range
of jE%j
Fermentation experiment
Blood without glucose Blood with glucose
Produced
ethanol (g/L),
mean ÆSD
Range of jE%j
(Eqs. (1B) and (2))
Produced ethanol
(g/L), mean ÆSD
Range of jE%j
(Eqs. (1B) and (2))
1 E. coli, E. faecalis 0.64 0.48–0.72 2–26 0.17 Æ0.06 26
a
0.69 Æ0.10 27–40
b
2 E. faecalis 2.94 0.71–0.97 67–76 0.76 Æ0.11 0–40, 1–27 1.05 Æ0.12 31–40, 3–35
3 K. pneumoniaea,
E. faecalis
0.33 0.55–1.20 63–261 0 – 0 –
4 Klebsiella spp.,
E. faecalis
0.84 0.65–1.15 4–37 0 – 0 –
5 E. faecalis 0 0 – 0 – 0 –
6 ND 0 0 – 0 – 0 –
7 ND 0 0 – 0 – 0 –
Abbreviation: ND, none (microbe) detected.
a
This value represents the best score achieved by applying the simplest model described by Eq. (1B).
b
Values resulted for the simplest model described by Eq. (1B).
V.A. Boumba et al. / Forensic Science International 232 (2013) 191–198 197
model in achieving the goal is different (due to the different
postmortem conditions, resulting in different alcohol patterns);
thus, one ‘‘tool’’ could be more effective and accurate than the
others for each case. Incidentally, all the proposed so far models,
these presented herein, as well as, those presented in our previous
relevant study [16], were applied successfully for 57 out of 60 cases
presented in Table 2. Nevertheless, lack of adequate evidence in
respect to the origin of ethanol for these cases does not allow a
definite conclusion, although putrefaction and extensive trauma of
the body are recognized aggravating factors for ethanol production
[3,4,21,25,26].
4. Concluding remarks
The present approach, approximates the problem of microbial
ethanol production with a novel, simple methodology. The models
are suggested as tools which could make feasible the quantification
of the microbial neo-formed ethanol in postmortem cases. The
alcohols 1-propanol and 1-butanol are proven consistent with
microbial production of ethanol and the two most significant
alcohols in the modeling process. Nevertheless, given the
complexity of the human body decomposition and the co-existing
microbial activity, our results point out that the more models are
available the more possible is one of them to estimate adequately
the postmortem ethanol in a given case. Of course, since the extent
of ethanol production in each case cannot be defined, we cannot
conclude, yet, which is the most suitable model for each case. Thus,
ongoing work should be conducted before suggesting the adequate
and reliable application of any model to real cases.
Acknowledgment
We thank Professor Stathis Frilligos for critical reading of the
manuscript.
References
[1] J.E.L. Corry, Possible sources of ethanol ante- and postmortem: its relation to the
biochemistry and microbiology of decomposition, J. Appl. Bacteriol. 44 (1978) 1–
56.
[2] C.L. O’Neal, A. Poklis, Postmortemproduction of ethanol and factors that influence
interpretation: a critical review, Am. J. Forensic Med. Pathol. 17 (1996) 8–20.
[3] K. Ziavrou, V.A. Boumba, T. Vougiouklakis, Insights into the origin of postmortem
ethanol, Int. J. Toxicol. 24 (2005) 69–77.
[4] F.C. Kugelberg, A.W. Jones, Interpreting results of ethanol analysis in postmortem
specimens: a review of the literature, Forensic Sci. Int. 165 (2007) 10–29.
[5] S.M. Petkovic´ , M.A. Simic´ , D.N. Vujic´ , Postmortem production of ethanol in
different tissues under controlled experimental conditions, J. Forensic Sci. 50
(2005) 204–208.
[6] R. Nanikawa, K. Ameno, Y. Hashimoto, K. Hamada, Medicolegal studies on alcohol
detected in dead bodies – alcohol levels in skeletal muscle, Forensic Sci. Int. 20
(1982) 133–140.
[7] W. Gubala, n-Butanol in blood as the indicator of how long a dead body lay in
water, Forensic Sci. Int. 46 (1990) 127–128.
[8] S. Felby, E. Nielsen, Congener production in blood samples during preparation and
storage, Blutalkohol 32 (1995) 5–58.
[9] F. Moriya, Y. Hashimoto, Postmortemproduction of ethanol and n-propanol in the
brain of drowned persons, Am. J. Forensic Med. Pathol. 25 (2004) 131–133.
[10] J. Furumiya, H. Nishimura, A. Nakanishi, Y. Hashimoto, Postmortem endogenous
ethanol production and diffusion from the lung due to aspiration of wood chip
dust in the work place, Leg. Med. (Tokyo) 13 (2011) 210–212.
[11] B.L. Soderberg, R.O. Salem, C.A. Best, J.E. Cluette-Brown, M. Laposata, Fatty acid
ethyl esters. Ethanol metabolites that reflect ethanol intake, Am. J. Clin. Pathol.
119 (2003) S94–S99.
[12] S. Arado´ ttir, S. Seidl, F.M. Wurst, B.A. Jo¨ nsson, C. Alling, Phosphatidylethanol in
human organs and blood: a study on autopsy material and influences by storage
conditions, Alcohol Clin. Exp. Res. 28 (2004) 1718–1723.
[13] R.D. Johnson, R.J. Lewis, D.V. Canfield, K.M. Dubowski, C.L. Blank, Utilizing the
urinary 5-HTOL/5-HIAA ratio to determine ethanol origin in civil aviation acci-
dent victims, J. Forensic Sci. 50 (2005) 670–675.
[14] G. Høiseth, R. Karinen, A.S. Christophersen, L. Olsen, P.T. Normann, J. Mørland, A
study of ethyl glucuronide in post-mortem blood as a marker of ante-mortem
ingestion of alcohol, Forensic Sci. Int. 165 (2007) 41–45.
[15] B.M. Appenzeller, M. Schuman, R. Wennig, Was a child poisoned by ethanol?
Discrimination between ante-mortemconsumption and post-mortemformation,
Int. J. Legal Med. 122 (2008) 429–434.
[16] V.A. Boumba, V. Economou, N. Kourkoumelis, P. Gousia, C. Papadopoulou, T.
Vougiouklakis, Microbial ethanol production: experimental study and multivari-
ate evaluation, Forensic Sci. Int. 215 (2012) 189–198.
[17] K.J. Ryan, C.G. Ray (Eds.), Sherris Medical Microbiology, 4th ed., McGraw Hill,
Norwalk, CT, 2004.
[18] A.W. Jones, L. Hyle´ n, E. Svensson, A. Helander, Storage of specimens at 4 8C or
addition of sodiumfluoride (1%) prevents formation of ethanol inurine inoculated
with Candida albicans, J. Anal. Toxicol. 23 (1999) 333–336.
[19] P. Saukko, B. Knight, The pathophysiology of death, in: Knight’s Forensic Pathol-
ogy, Oxford University Press Inc., New York, 2004, pp. 52–97 (Chapter 2).
[20] V.A. Boumba, K. Ziavrou, T. Vougiouklakis, Biochemical pathways generating
post-mortem volatile compounds co-detected during forensic ethanol analysis,
Forensic Sci. Int. 174 (2008) 133–151.
[21] R. Hanzlick, Ethanol concentration in decomposing bodies: another look, less
concern, Am. J. Forensic Med. Pathol. 30 (2009) 88–89.
[22] R.E. Zumwalt, R.O. Bost, I. Sunshine, Evaluation of ethanol concentrations in
decomposed bodies, J. Forensic Sci. 27 (1982) 549–554.
[23] J.A. Morris, L.M. Harrison, S.M. Partridge, Post mortembacteriology: a re-evalua-
tion, J. Clin. Pathol. 59 (1) (2006) 1–9.
[24] H.A. Sulkowski, A.H. Wu, Y.S. McCarter, In-vitro production of ethanol in urine by
fermentation, J. Forensic Sci. 40 (1995) 990–993.
[25] D.M. Butzbach, The influence of putrefaction and sample storage on post-mortem
toxicology results, Forensic Sci. Med. Pathol. 6 (2010) 35–45.
[26] G. Scopp, Postmortem toxicology, Forensic Sci. Med. Pathol. 2 (2010) 314–325.
V.A. Boumba et al. / Forensic Science International 232 (2013) 191–198 198