Bioprocess and Biosystems Engineering (2018) 41:555–564

https://doi.org/10.1007/s00449-018-1890-7

RESEARCH PAPER

Pilot-scale process development for low-cost production

of a thermostable biodiesel refining enzyme in Escherichia coli
Florencia Eberhardt1 · Andres Aguirre1 · Luciana Paoletti1 · Guillermo Hails1 · Mauricio Braia1 · Pablo Ravasi1 ·
Salvador Peiru1,2 · Hugo G. Menzella1,2

Received: 26 October 2017 / Accepted: 31 December 2017 / Published online: 10 January 2018
© Springer-Verlag GmbH Germany, part of Springer Nature 2018

Abstract
Biodiesels produced from vegetable oils have a major quality problem due to the presence of steryl glucosides (SGs), which
form precipitates that clog filters and cause engine failures. Recently, we described an enzymatic process for removing SGs
from biodiesel. However, industrial adoption of this technology was hindered by the cost of the steryl glucosidase (SGase)
enzyme used. Here we report the development and validation at the pilot scale of a cost-efficient process for manufacturing
the SGase. First, we tested various low-cost carbon sources for the Escherichia coli producing strain, ultimately develop-
ing a fed-batch fermentation process that utilizes crude glycerol as a feedstock. Next, we designed an efficient process for
isolating the SGase. That process uses a novel thermolysis approach in the presence of a non-ionic detergent, centrifugation
to separate the solids, and ultrafiltration to concentrate and formulate the final product. Our cost analysis indicates that on a
large scale, the dose of enzyme required to eliminate SGs from each ton of biodiesel will have a manufacturing cost below
$1. The new process for manufacturing the SGase, which will lead to biodiesels of a higher quality, should contribute to
facilitate the global adoption of this renewable fuel. Our technology could also be used to manufacture other thermostable
proteins in E. coli.

Keywords Bioprocess development · Microbial thermolysis · Biofuels

Introduction alternative fuel is affected by the presence of contaminants that

form sediments and cause the failure of engines. Steryl gluco-
Biodiesels are renewable fuels mostly produced from veg- sides (SGs), present in different biodiesels at concentrations
etable oils, including those from soybean, palm, sunflower, ranging from 10 to 300 ppm [3, 4], have been identified as the
rapeseed, jatropha, and others. Currently, biodiesel mandates major component of such sediments [5–7]. Thus, the selective
are set in more than 60 countries, many of which belong to removal of SGs could produce biodiesels of a superior quality,
high-consuming regions like the European Union and North increasing the likelihood that these renewable fuels will be
America [1]. Global biodiesel production is projected to con- adopted by consumers. Currently, the only available method
tinue its rapid increase and reach more than 40 billion liters capable of completely removing SGs from biodiesel is distilla-
by 2020 [2]. Unfortunately, the acceptance of biodiesel as an tion, an energy-intensive and expensive process [8, 9].
Recently, we described an efficient method for remov-
Electronic supplementary material The online version of this ing SGs from biodiesel. The method is based on the use of
article (https://doi.org/10.1007/s00449-018-1890-7) contains enzymes with steryl glycosidase (SGase) activity. The most
supplementary material, which is available to authorized users. efficient SGase tested so far is a thermostable β-glucosidase
from Thermococcus litoralis. Because this hyperthermo-
* Hugo G. Menzella
hmenzella@fbioyf.unr.edu.ar philic archaeon is difficult to cultivate, a synthetic codon-
optimized version of the SGase gene was designed and
1
Genetic Engineering and Fermentation Technology successfully expressed in E. coli [10]. The producing strain
Laboratory. Facultad de Ciencias Bioquímicas y was then optimized through extensive engineering, which
Farmacéuticas, Universidad Nacional de Rosario-Conicet,
Suipacha 531, 2000 Rosario, Argentina included the use of different promoters and the co-expres-
2 sion of molecular chaperones to significantly increase the
Keclon S.A. Tucuman 7180, 2000 Rosario, Argentina

13
Vol.:(0123456789)
556 Bioprocess and Biosystems Engineering (2018) 41:555–564

production of the SGase. Additionally, a high cell density, growth was determined by measuring optical density at
fed-batch fermentation process has been developed [11]. 600 nm ­(OD600) in a General Electric NovaspecIII spec-
However, much remains to be done to achieve a cost-effec- trophotometer. HM medium consists of 20.8 g/L K ­ H2PO4,
tive process for manufacturing SGase. 3 g/L ­(NH4)2PO4, 3.25 g/L K(OH), and 4 g/L ­NaH2PO4.
Currently, industrial enzymes produced by fermentation Microplates (Thermo Fisher Scientific) were prepared by
are considered commodities. The carbon source used in the adding to each well, 190 μL of HM medium supplemented
fermentation processes can account for up to 50% of the total with 1% of the source of carbon being tested and the neces-
manufacturing cost [12, 13]. In addition, downstream pro- sary antibiotics. Each well was inoculated with 10 μL of
cesses involving a few simple operations can greatly reduce the centrifuged culture. The microplates were incubated
capital expenditures and operating costs [14–17]. In this at 37 °C for 12 h with continuous shaking in a microplate
work, we analyze the impact of different carbon sources and reader (Synergy HT). ­OD600 were recorded every 30 min.
the post-induction feeding strategy on the productivity of the
SGase in E. coli, aiming to develop a commercial fermen- Growth conditions for expressing the SGase
tation process for producing this enzyme on a large scale.
Additionally, a simple thermolysis method is described for Recombinant E. coli strains were grown overnight at 37 °C
facilitating the recovery of the enzyme. The results described in 2 mL of LB medium supplemented with the appropriate
here could be used to design processes for producing other antibiotic.
thermostable enzymes in E. coli in a cost-effective manner. To test the expression of the SGase, 5 mL of HM medium
was supplemented with different carbon sources (glycerol,
crude glycerol, glucose, sucrose, or molasses) at 1% (w/v)
Materials and methods and was inoculated with an overnight culture to an initial
­OD600 = 0.1. Following incubation at 200 rpm and 37 °C,
General protein expression was induced at O­ D600 = 1 with 0.4 g/L of
L-arabinose. Growth was allowed to continue at 37 °C for
Enzymes were obtained from New England Biolabs (USA) 6 additional hours.
and used as recommended. E. coli Top10 (Invitrogen) was
used for plasmid propagation during cloning steps. E. coli Cell disruption
BL21(DE3)AI served as the expression host. E. coli strains
were made chemically competent with a kit from Zymo Cells and supernatants of the cultures were separated by
Research (USA). The concentrations of kanamycin and centrifugation. The cell pellets were resuspended in 20 mM
chloramphenicol used were 50 and 20 mg/L, respectively. citrate (pH 6.0) and 20 mM NaCl, normalized to a final
­OD600 = 4 per mL. Cells were disrupted on ice in a GEX
Insertion of a cassette for metabolizing sucrose 600 Ultrasonic Processor.

The cscAKB operon was amplified by PCR using oligonu- Activity assays
cleotides that match the 5′ and 3′ ends of the operon. ScaI
restriction sites were added at the end of both oligonucleo- Expression of the SGase under different growth condi-
tides, and the resulting DNA fragments were inserted into tions was determined by measuring β-Glucosidase activity,
identical sites of the plasmid pACYC184. The resulting according to a method previously developed in our labora-
plasmid was digested with BclI and SacII, and the fragment tory [18]. One unit (U) was defined as the amount of enzyme
containing the operon and the Cm resistance cassette was required for the hydrolysis of 1 μmol of pNPG per min,
inserted into the lacZ locus of E. coli BL21(DE3)AI, as under the described assay conditions.
described by Datzenko et al. [20].
High cell density fermentation
Growth conditions for expression studies in 96‑well
microplates Seed cultures of E. coli BL21(DE3)AI and E. coli NK5,
harboring the plasmids pKCN-BAD-SGase and pGro7 [11],
From stocks stored at − 80 °C, strains were grown over- were prepared in 1-L Erlenmeyer flasks containing 0.1 L of
night on Luria–Bertani (LB) agar plates at 37 °C. Individual HM medium. The medium was supplemented with glycerol,
colonies were used to inoculate 5 mL of HM medium in crude glycerol, glucose, sucrose, or molasses as the carbon
50 mL glass tubes, the medium being supplemented with source. The cultures were grown at 37 °C in an incubator
the appropriate carbon source. The cultures were incubated with shaking at 200 rpm.
overnight at 37 °C with shaking at 200 rpm. Microbial

13
Bioprocess and Biosystems Engineering (2018) 41:555–564 557

Fed-batch fermentations were conducted in a 3-L bioreac- Table 1  Growth of BL21(DE3)AI/pKCN-BAD-SGase pGro7 and
tor (New Brunswick BioFlo 115, USA) containing 1 L of the NK5/pKCN-BAD-SGase pGro7 strains with different carbon sources
same medium. The temperature and stirring were maintained Carbon source BL21(DE3)AI/pKCN- NK5/pKCN-
at 37 °C and 1200 rpm, respectively. The pH was main- BAD-SGase BAD-SGase
tained at 7.0 by addition of 28% N
­ H4OH. The concentration pGro7 pGro7
of dissolved oxygen was maintained at 30% of saturation. Final ­OD600nm
The feeding process was initiated when the carbon source Glucose 2.65 ± 0.19 2.75 ± 0.10
initially present in the medium was exhausted. A feeding Glycerol 2.65 ± 0.13 2.46 ± 0.18
solution containing 600 g/L of glycerol, crude glycerol, glu- Crude glycerol 2.68 ± 0.25 2.49 ± 0.17
cose, sucrose, or molasses, plus 20 g/L ­MgSO4·7H2O, was Sucrose 0.48 ± 0.19 2.67 ± 0.09
added by following an exponential feeding rate until protein Molasses 1.35 ± 0.14 3.08 ± 0.25
expression was induced. µ ­(h−1)
The feeding rate (F, mL/h) was determined by Eq. (1) Glucose 0.63 ± 0.09 0.62 ± 0.05
[19], to maintain the specific growth rate at 0.25 1/h: Glycerol 0.55 ± 0.07 0.54 ± 0.09
F = X0 V0 e(u.t)∕ S0 YX∕S Crude glycerol 0.59 ± 0.07 0.57 ± 0.08
Sucrosa – 0.52 ± 0.07
X0 is the biomass concentration (g/L) when the feeding is Molasses – 0.55 ± 0.06
started. V0 is the initial volume (L). u is the desired specific
growth rate (1/h). S0 is the carbon source concentration in Growth parameters are shown for the strains BL21(DE3)AI and NK5,
the feeding solution (g/L). YX/S is the substrate yield. when both harbor the plasmids pGro7 and pKCN-BAD-SGase. The
strains were cultivated in 96-well plates with a minimal medium. The
When the ­OD600 reached 100, expression of the SGase experimental error is the standard deviation for n = 3
was induced by adding L-arabinose at a final concentration
of 0.4 g/L. Afterwards, the feeding rate was maintained at
8 g/h, except in the post-induction optimization experiments. per metric ton. To explore lower-cost alternatives (below
Pilot-scale fermentations were conducted at 1300 L. The $200 per ton), we decided to test sucrose, glycerol, and
culture medium, relative glycerol feeding rate, pH, and tem- industrial products containing these compounds—such
perature were those used for the laboratory-scale fermenta- as molasses and biodiesel-derived crude glycerol—as
tions. The inoculum was prepared in a 50 L bioreactor, to feedstocks for the production of SGase. While derivatives
obtain an inoculum of the same relative size as that in the of E. coli BL21 can grow efficiently by utilizing glucose
laboratory-scale experiments. After the fermentation, cells or glycerol, these strains cannot use sucrose as a carbon
were disrupted on ice at 1000 bars through a high-pressure source, because they lack the genes for transporting and
homogenizer (GEA NiroSoavi, Panda Plus 2000), except in metabolizing this disaccharide. E. coli W strains grow
the thermolysis experiments. with sucrose as the sole carbon and energy source using
the cscB, cscA, and cscK genes, which mediate the trans-
Thermolysis assays port and hydrolysis of sucrose and the phosphorylation of
fructose, respectively. Therefore, we first engineered the
High cell density cultures were incubated with none, 0.2, BL21-based producing strain by inserting a sucrose utili-
or 2% of the indicated detergent, and the mixtures were zation cassette containing these three genes into the lacZ
heated at 70 °C for 30 min. Then, the mixtures were cen- locus of the chromosome. For this step, the operon con-
trifuged at 10,000 rpm for 20 min in a bench centrifuge taining the genes and the native promoter was amplified by
(Eppendorf 5804R). Purified samples were analyzed by PCR using oligonucleotides with extensions homologous
SDS–PAGE. Concentrations of the SGase were estimated to the target sequence in the chromosome. The operon was
by β-glucosidase activity. inserted using the Lambda red system described by Dat-
senko et al. [20]. The engineered strain was named NK5
and was tested along with the native strain for its ability to
Results and discussion grow on sucrose, glucose, or glycerol supplementing mini-
mal medium. For each strain, three independent clones
Analysis of different carbon sources were assessed.
for the production of the SGase Table 1 shows that, as expected, the native BL21-derived
and the NK5 strains grew well on glucose and glycerol. Only
Bioprocesses based on E. coli mostly utilize for their car- the NK5 strain, which harbors the cassette for metaboliz-
bon and energy source glucose obtained from the hydroly- ing sucrose, could grow on sucrose. The strains also grew
sis of cornstarch. This glucose has an average cost of $800 on crude glycerol at a rate comparable to that obtained

13
558 Bioprocess and Biosystems Engineering (2018) 41:555–564

for refined glycerol. Likewise, the NK5 strain exhibited Figure 1 summarizes the results. In all of the cases, the
a growth profile in molasses similar to the one shown for maximum ­OD600 and SGase activity were obtained after 25 h
sucrose. The minor growth observed for the BL21 strain in of cultivation. The highest biomass concentration (160 OD
molasses is probably due to a small amount of glucose being units) and SGase production (280 units/mL) were obtained
naturally present in this feedstock. for refined and crude glycerol. This result is consistent with
Next, shake flask cultures were grown in minimal medium multiple reports indicating that glycerol, contrary to glucose
supplemented with either 10 g/L of glucose, 11.5 g/L of and sucrose, is an attractive feedstock for use in E. coli fed-
refined or crude glycerol (containing 87% of glycerol), batch fermentations, because acetate is not produced. The
10 g/L of sucrose, or 16.7 g/L of molasses (with a sucrose accumulation of this byproduct has been widely reported
concentration of 60%), to analyze the impact of each car- to be the major factor affecting both cell growth and the
bon source on the level of expression of the SGase. Table 2 production of recombinant proteins in E. coli in this type of
shows the results. In all of the cases, a similar final ­OD600 fermentation [19, 22–24].
was reached, confirming that all of the tested carbon sources Bruschi and co-workers obtained identical yields when
efficiently support the growth of the producing strains. The producing DAMP4, a surfactant peptide from either glu-
specific production of the SGase was comparable for all of cose or sucrose in an engineered strain, when the same set
the feedstocks, being about 10% lower for refined glycerol of genes for sucrose metabolism used here were expressed
and sucrose. from a multicopy plasmid [25]. In our experiments, sucrose
yielded a lower final biomass and less SGase activity than
Fermentation process development glucose did. However, in the fed-batch fermentations using
sucrose or molasses as a carbon source, we detected the
To test the efficacy of the different carbon sources in the accumulation of sucrose (data not shown). This accumula-
high cell density cultures typically used for the industrial tion might account for the poorer performance. A possible
production of enzymes, we developed fed-batch processes explanation is a gene dose effect, in which the integration of
using glucose, refined and crude glycerol, sucrose, or molas- a single copy of the genes involved in sucrose metabolism
ses. In all of the cases, a 3-L lab scale bioreactor was used. into the chromosome might cause insufficient protein expres-
The recombinant strains were grown at 37 °C in 1 L of HM sion, compared to expression from multicopy plasmids.
medium supplemented with 20 g/L of the corresponding Our results and the economic advantage of using a waste
carbon source. The pH was maintained at 6.9 by the addi- product induced us to choose crude glycerol as the preferred
tion of 28% N ­ H4OH. The concentration of dissolved oxygen carbon source for producing the SGase. All further experi-
was kept above 30% of saturation. After the initial amount ments were conducted with this feedstock.
of a carbon source was exhausted, cultures were fed with a
solution of 60% of the corresponding substrate. A feeding Effect of post‑induction strategies for feeding crude
strategy using a balance mass equation was used to main- glycerol on the production of the SGase
tain the specific growth rate at 0.25 [19, 21]. When cultures
reached an O ­ D600 of 100, 0.4 g/L of L-arabinose was added The post-induction strategy for feeding nutrients greatly
to induce the production of the SGase. The feeding rate was impacts cell growth and protein production [26, 27]. There-
kept constant at a rate of 8 g/h of substrate per L. fore, we further optimized our process by comparing four

Table 2  Expression of the Carbon source Glucose Glycerol Crude Glycerol Sucrose Molasses
SGase in shake flask cultures
OD600nm
BL21(DE3)AI/pKCN-BAD- 6.07 ± 0.23 5.14 ± 0.11 5.23 ± 0.24 – –
SGase pGro7
NK5/pKCN-BAD-SGase 5.95 ± 0.11 5.29 ± 0.22 5.91 ± 0.31 5.11 ± 0.10 5.26 ± 0.08
pGro7
SGase Activity (U/DO)*
BL21(DE3)AI/pKCN-BAD- 3.35 ± 0.38 2.98 ± 0.29 3.41 ± 0.32 – –
SGase pGro7
NK5/pKCN-BAD-SGase 3.22 ± 0.31 3.07 ± 0.21 3.12 ± 0.15 2.96 ± 0.21 3.41 ± 0.40
pGro7

Strains BL21(DE3)AI and NK5 harboring pGro7 and pKCN-BAD-SGase, were cultivated in shake flasks
in HM medium with different carbon sources. The activity of the SGase was analyzed. The experimental
error is the standard deviation for n = 3

13
Bioprocess and Biosystems Engineering (2018) 41:555–564 559

Fig. 1  Fed-batch fermentations using different carbon sources. strains. NK5 harbors the plasmids pKCN-BAD-SGase and pGro7.
Growth curves of fed-batch cultures in minimal medium with a Fermentations were conducted in duplicate. The figure shows the
glucose, b glycerol, c crude glycerol, d sucrose, and e molasses as average of the duplicate runs. In all of the cases, the data varied by
the sole carbon source for the BL21(DE3)AI (a–c) and NK5 (d–e) less than 10%

different feeding strategies: (1) keeping the post-induction Table 3 shows the results. The three alternative post-
feeding rate of crude glycerol constant at 8 or (2) 12 g/L.h, induction feeding strategies yielded higher final activities
(3) a strategy in which the feeding started at 8 g/L.h and was of the SGase than did the original condition of feeding glyc-
linearly increased at a rate of 0.56 g/L.h, and (4) a feeding erol at a constant rate of 8 g/L h. Moreover, the duration of
profile in which the initial rate of 10 g/L.h was increased the fermentations decreased. A maximum productivity of
based on the culture ­OD600 (0.1 g/L.h per OD unit). In all of 16.2 U/mL.h was achieved with a constant feeding rate of
the cases, samples were analyzed hourly. The cultures were 12 g/L h, 50% higher than the productivity of the original
harvested when the SGase activity started to decline. protocol.

13
560 Bioprocess and Biosystems Engineering (2018) 41:555–564

Table 3  Effect on the SGase (U/mL) Final ­OD600 Fermentation pro- Productiv-
productivity of the SGase cess time (h) ity (U/ml h)
of different post-induction
strategies for feeding crude Constant fed 8 g/l/h 280 137 26 10.8
glycerol
Constant fed 12 g/l/h 405 140 25 16.2
Linear fed increase 390 146 25 15.6
Fed proportional to O­ D600 355 133 25 14.2

Four strategies were tested: constant feed rates at 8 or 12 g/L h; a linearly increasing feed rate at 0.7 g/L.h;
and an O­ D600-based increasing feed rate at 0.1 g/L h per O ­ D600 unit. Fermentations were conducted in
duplicate. In all of the cases, the data varied by less than 10%

Different feeding strategies have been employed by vari- spaced Rushton turbines that have six blades and four dia-
ous authors for the post-induction phase of recombinant metrically opposed baffles (Supplementary Fig. 1). An
cultures. The strategies include the following: a constant inoculum bioreactor was used as a step between shake flask
feeding rate [28–30]; a feeding rate that changes linearly cultures and the pilot-scale bioreactor, so that the inoculum
[27]; a feeding rate that changes exponentially, beginning for the latter would have the same high concentration of
during the pre-induction phase [31, 32]; and a feeding rate cells and relative volume. The components of the medium,
with feedback control that resembles a DO-stat or a pH-stat relative glycerol feeding rate, and pH and temperature of the
[33]. In all of these cases, the productivity varied widely, process were the same as those used for the laboratory-scale
such that a “universal” post-induction feeding profile could fermentations.
not be established. Based on the data available in the litera- Figure 2 and Table 4 show the growth and the SGase
ture, it seems that optimization is required for each new bio- activity for the pilot-scale fermentation. The final cell den-
process in E. coli. In our case, the productivity of the SGase sity, substrate yield, and SGase productivity were similar to
increased with the amount of glycerol fed. However, further those obtained in the laboratory-scale fermentations. Glyc-
increments in the feeding rate were not tested, because a erol did not accumulate, indicating that the substrate was
higher feeding rate would create an oxygen demand that efficiently consumed throughout the fermentation. Values of
would be difficult to satisfy in large-scale fermentations. dissolved oxygen tension reached using 1 VVM of air were
also similar for the laboratory and pilot-scale fermentations,
Scale‑up of the fermentation process indicating that the working conditions were successful in
achieving similar oxygen transfer rates, one of the major
To validate our laboratory-scale process at the pilot scale, limitations when scaling-up fermentation processes [37].
we conducted fermentations in a 1300 L fermenter using Scaling up E. coli fermentation processes from the labo-
a constant impeller tip speed (7 m/s) as a scale-up criteria ratory to the pilot plant often lowers yields due to various
[34–36]. Because of its simplicity, this is one of the most factors, including insufficient oxygen transfer [38], plasmid
used scale-up methods. The pilot-scale tank had a height/ loss [39], and acetate formation from high local concentra-
diameter ratio of 2.8 and was equipped with three equally tions of glucose caused by insufficient mixing [40–42].

Fig. 2  Fed-batch fermenta-

tion of BL21(DE3)AI strain
harboring the pKCN-BAD-
SGase pGro7 plasmid at a
1300-L scale. Time courses of
­OD600 nm (blue diamond), SGase
activity (red square), flow rate
of glycerol (green triangle), and
dissolved oxygen (gray circle)
are shown. The fermentations
were conducted in duplicate.
The figure shows the average of
the duplicate runs. In all of the
cases, the data varied by less
than 10%

13
Bioprocess and Biosystems Engineering (2018) 41:555–564 561

Table 4  Comparison of fed-batch fermentations at the laboratory and The disruption of cells by mechanical methods, however,
pilot plant scales presents several problems: (1) all soluble cellular proteins
Scale 1300 l 3l are released, which renders subsequent purification difficult
and more expensive; (2) mechanical disruption results in
Final ­OD600 140 150
the micronization of cell debris and increases viscosity by
Residual glycerol (g/l) 0 0
releasing DNA, which reduces the efficiency of particulate
Y/S (dry cell weight/glycerol consumed) 0.41 0.40
removal during subsequent centrifugation or membrane fil-
SGase activity (U/ml) 410 400
tration; and (3) the required machines have a high capital
Fermentation process time (h) 25 25
cost.
Productivity (U/ml/h) 16.4 16
Since the SGase is a thermostable protein, we decided to
Fermentations of E. coli BL21(DE3)AI strain harboring the pKCN- evaluate a thermolysis approach to isolate it. First, we tested
BAD-SGase pGro7 plasmid were conducted in duplicate. The table the effect of heat on the release of the product, by raising
shows the average of the duplicate runs. In all of the cases, the data the temperature of the fermentation broth directly at the end
varied by less than 10%
of the fermentation at the 1300 L scale. The activity and
the purity of the SGase in the supernatant were analyzed.
In our pilot-scale experiments, the SGase activity was Unfortunately, these initial tests showed no activity recov-
slightly higher than what we expected, when we considered ered from the supernatant. In all of the cases, SDS PAGE
the data from our small-scale fermentations. The yield coef- analysis revealed that the SGase remained associated with
ficient YX/S was close to the theoretical value for glycerol the insoluble fraction after the heat treatment (Fig. 3a).
(0.45) and similar to the yields obtained at the laboratory- The SGase is highly homologous to BGPh from Pyrococ-
scale, indicating that the scale-up criteria adopted for the cus horikoshii, which is associated with the membrane in
bioprocess were successful. Taken together, these results the native organism [45]. This information suggested that
suggest that further scale up to 25 m3 bioreactors for the the SGase might associate with the E. coli membrane after
commercial production of the SGase is feasible. cells are disrupted by heat and that adding a detergent could
help to release the enzyme.
Isolation of the SGase by thermolysis The subsequent experiment confirmed that addition of
0.2% Triton X-100 to the fermentation broth releases the
E. coli has a major disadvantage as a host for producing enzyme from the insoluble fraction (Fig. 3b). However, the
proteins, in that this bacterium does not efficiently release high cost of Tritron X-100 prevents its use for producing the
the product to the medium. Thus, methods for releasing SGase at an industrial scale. Therefore, we sought to iden-
recombinant proteins from cells following fermentation tify a cost-effective substitute for Triton X-100. We repeated
are required. High-pressure homogenizers and bead mills the thermolysis procedure, again by incubating the broth at
have typically been used at the industrial scale [43, 44]. 70 °C for 30 min, but in the presence of various ionic and

Fig. 3  Isolation of the SGase by thermolysis. a SDS PAGE of cell tion without centrifugation (lane 1), or by thermolysis in the presence
extracts obtained from a fermentation, by high-pressure homog- of 2, 0.2%, or no Triton X-100 (lanes 2, 3 and 4, respectively) fol-
enization without centrifugation (line 1), or by thermolysis without lowed by centrifugation. The figure shows the average values of three
detergent followed by centrifugation (lanes 2–3). b SDS-PAGE of experiments that have standard deviation less than 10%. P pellet, S
cell extracts obtained a fermentation, by high-pressure homogeniza- supernatant

13
562 Bioprocess and Biosystems Engineering (2018) 41:555–564

non-ionic detergents. Figure 4 shows that all of the non-ionic to temperatures above 70 °C. Recently, an effective ther-
detergents tested resulted in a high recovery of the SGase. molysis procedure was described to isolate peptides fused
0.2% lauryl alcohol ethoxylate (7 mol) led to a recovery to a thermostable protein [47]. In addition to reduced capital
of the SGase in the soluble fraction that was greater than expenditures and simplified recovery of the target protein,
90%. This detergent was chosen for use at the industrial scale thermolysis has further advantages for use at the industrial
because of its low toxicity and cost. The addition of ionic scale, including killing the host and deactivating proteases.
detergents abolished nearly all activity, presumably by dena- The procedure described here to recover the SGase is,
turing the SGase. After thermolysis, continuous centrifuga- to the best of our knowledge, the first one combining ther-
tion easily removed the solids. 10K hollow-fiber cartridges molysis with the use of a non-ionic detergent. The protocol
concentrated the resulting supernatant five-fold, producing a might serve to purify thermostable proteins associated with
final SGase preparation that was stable at room temperature membranes or that are prone to aggregate, because hydro-
(data nor shown). phobic interactions are favored at the high temperatures used
Several reports indicate that incubation of E. coli above to disrupt cells.
60 °C releases cytoplasmic proteins [16, 17]. Watson et al.
[46] and Ren et al. [15] reported the release of cytoplas- Cost analysis
mic proteins from E. coli within ten minutes of exposure
Previously, we demonstrated that 260,000 units of the SGase
(which is equivalent to 7 g of the enzyme) is sufficient to
completely remove the average amount of SGs (100 ppm)
present in one metric ton of soybean-derived biodiesel [10,
11]. The process illustrated in in Fig. 5 consist of a high cell
density, fed-batch fermentation, a thermolysis step in the
presence of a non-ionic detergent added to the fermenter
followed by solids removal by continuous centrifugation and
ultrafiltration to concentrate and formulate the final product.
At 1300 L scale, and considering an efficiency of recov-
ery of 70%, approximately one liter of culture could provide
the amount of enzyme required to treat one ton of biodiesel.
Based on this calculation, and assuming a similar yield, we
estimated the cost of producing the SGase at the 25,000 L
scale, using crude glycerol as a feedstock and thermolysis
Fig. 4  Recovery of the SGase by thermolysis in the presence of dif- followed by centrifugation and ultrafiltration to recover the
ferent detergents. SGase activity was measured after thermolysis in enzyme. Figure 6 and Supplementary File 1 show the results.
the presence of 0.2% of each detergent. NPEt 10M: nonylphenol eth- The final cost per dose is estimated to be $0.71.
oxylate 10 mol, LAEt 7M: lauryl alcohol ethoxylate 7 mol, AEt com-
Raw materials include all of the ingredients of the cul-
mercial mixture of alcohols ethoxylate, SDS sodium dodecyl sulfate,
LSABS linear sodium alkylbenzene sulphonate, LAABS linear ammo- ture medium, where crude glycerol represents 50% of the
nium alkylbenzene sulfonate total cost in our case. Utilities comprise cooling and process

Fig. 5  Flow diagram for the

proposed process to produce
and purify the SGase. The pro-
cess includes high cell-densitiy
fermentation, thermolysis,
continuous centrifugation and
ultrafiltration

13
Bioprocess and Biosystems Engineering (2018) 41:555–564 563

We estimate that, using this process on a large scale, the dose

of enzyme required to eliminate SGs from biodiesel will
have a manufacturing cost below $1. This novel technology
should be attractive to biodiesel producers, by facilitating
the global adoption of the renewable fuel. In addition, the
low-cost process described here could be a general method
for producing other thermostable proteins in E. coli.

References
1. Lane J (2014) Biofuels Mandates Around the World: 2015
Fig. 6  Cost estimate for the production of the SGase. Values were 2. OCED F (2011) OECD-FAO agricultural outlook 2011–2020.
estimated for the process described in Fig. 5 at the 25,000 L scale, Organisation for Economic Co-operation and Development
using crude glycerol as a carbon source (OECD) Publishing, doi, Paris
3. Ringwald S (2007) Biodiesel characterization in the QC environ-
ment. The 98th AOCS annual meeting abstracts. AOCS Press,
Urbana
water, steam, and electricity. Facility-dependent costs
4. Tang H, De Guzman R, Salley S, Ng KS (2010) Comparing pro-
include depreciation and insurance. All of our costs were cess efficiency in reducing steryl glucosides in biodiesel. J Am Oil
estimated for a facility located in Argentina, one of the larg- Chem Soc 87:337–345
est producers of biodiesel in the world. 5. Bondioli P, Cortesi N, Mariani C (2008) Identification and quan-
tification of steryl glucosides in biodiesel. Eur J Lipid Sci Technol
Detailed cost analyses for producing industrial enzymes
110:120–126
are scarce in the public domain. Recently, Potvin and co- 6. Haagenson DM, Perleberg JR, Wiesenborn DP (2014) Fractiona-
workers presented a cost analysis for the large-scale produc- tion of canola biodiesel sediment for quantification of steryl glu-
tion of phytase in Pichia pastoris. They estimated a produc- cosides with HPLC/ELSD. J Am Oil Chem Soc 91:497–502
7. Pfalzgraf LLI, Foster J, Poppe G (2007) The effect of minor
tion cost equivalent to 5 cents per gram [48], lower than the
components on cloud point and filterability. Inf Suppl Biorenew
cost presented here, which is approximately 10 cents per Resour 4:17–21
gram. However, they used a semi-continuous fermentation 8. Camerlynck S, Chandler J, Hornby B, van Zuylen I (2012) FAME
at the 180,000 L scale, which due to longer production times filterability: understanding and solutions. SAE Int J Fuels Lubr
5:968–976
and reduced culture downtimes results in higher volumetric
9. Plata V, Haagenson D, Dağdelen A, Wiesenborn D, Kafarov V
productivities. Klein-Marcuschamer et al. analyzed how the (2015) Improvement of palm oil biodiesel filterability by adsorp-
cost of cellulases impacted ethanol production, finding that tion methods. J Am Oil Chem Soc 92:893–903
the breakdown of the production cost for the enzymes is 10. Peiru S, Aguirre A, Eberhardt F, Braia M, Cabrera R, Menzella
HG (2015) An industrial scale process for the enzymatic removal
similar to the one that we report here [49]. of steryl glucosides from biodiesel. Biotechnol Biofuels 8:223
In our case, the production of the SGase at the 25,000 L 11. Eberhardt F, Aguirre A, Menzella HG, Peiru S (2017) Strain engi-
scale could provide a highly affordable solution for elimi- neering and process optimization for enhancing the production
nating SGs that contaminate biodiesel. A facility with this of a thermostable steryl glucosidase in Escherichia coli. J Indus
Microbiol Biotechnol 44:141–147
capacity should provide sufficient enzyme to treat more than 12. Jiang G, Hill DJ, Kowalczuk M, Johnston B, Adamus G, Irorere
8 million tons of biodiesel, a value which represents 40% of V, Radecka I (2016) Carbon sources for polyhydroxyalkanoates
worldwide production in 2016 [50]. and an integrated biorefinery. Int J Mol Sci 17:1157
13. Rude MA, Schirmer A (2009) New microbial fuels: a biotech
perspective. Curr Opin Microbiol 12:274–281
14. Ren X, Yu D, Han S, Feng Y (2007) Thermolysis of recombinant
Conclusions Escherichia coli for recovering a thermostable enzyme. Biochem
Eng J 33:94–98
We have developed a cost-effective bioprocess for produc- 15. Ren X, Yu D, Yu L, Gao G, Han S, Feng Y (2007) A new study of
cell disruption to release recombinant thermostable enzyme from
ing a SGase in E. coli. The SGase is an enzyme capable of Escherichia coli by thermolysis. J Biotechnol 129:668–673
removing biodiesels derived from vegetable oils SGs, the 16. Balasundaram B, Harrison S, Bracewell DG (2009) Advances in
major contaminant present. The process involves a high cell product release strategies and impact on bioprocess design. Trends
density, fed-batch fermentation that uses crude glycerol as Biotechnol 27:477–485
17. Koschorreck K, Wahrendorff F, Biemann S, Jesse A, Urlacher VB
the carbon source,a waste product from biodiesel production. (2017) Cell thermolysis–A simple and fast approach for isolation
The fermentation is followed by thermolysis in the presence of bacterial laccases with potential to decolorize industrial dyes.
of a non-ionic detergent to release the enzyme from cells. Process Biochem 56:171–176
Centrifugation separates the solids from the supernatant. 18. Aguirre A, Peiru S, Eberhardt F, Vetcher L, Cabrera R, Men-
zella HG (2014) Enzymatic hydrolysis of steryl glucosides, major
Ultrafiltration concentrates and formulates the final product.

13
564 Bioprocess and Biosystems Engineering (2018) 41:555–564

contaminants of vegetable oil-derived biodiesel. Appl Microbiol 34. Junker BH (2004) Scale-up methodologies for Escherichia coli
Biotechnol 98:4033–4040 and yeast fermentation processes. J Biosci Bioeng 97:347–364
19. Lee SY (1996) High cell-density culture of Escherichia coli. 35. Ju L-K, Chase G (1992) Improved scale-up strategies of bioreac-
Trends Biotechnol 14:98–105 tors. Bioprocess Biosyst Eng 8:49–53
20. Datsenko KA, Wanner BL (2000) One-step inactivation of chro- 36. Paul EL, Atiemo-Obeng VA, Kresta SM (2004) Handbook of
mosomal genes in Escherichia coli K-12 using PCR products. industrial mixing: science and practice. John Wiley & Sons
Proceedings of the National Academy of Sciences 97, 6640–6645 37. Garcia-Ochoa F, Gomez E (2009) Bioreactor scale-up and oxygen
21. Kodumal SJ, Patel KG, Reid R, Menzella HG, Welch M, Santi transfer rate in microbial processes: an overview. Biotechnol Adv
DV (2004) Total synthesis of long DNA sequences: synthesis of a 27:153–176
contiguous 32-kb polyketide synthase gene cluster. Proceedings of 38. Holmberg A, Ranta J (1982) Procedures for parameter and state
the National Academy of Sciences of the United States of America estimation of microbial growth process models. Automatica
101, 15573–15578 18:181–193
22. Yee L, Blanch H (1993) Recombinant trypsin production in high 39. O’Sullivan LM, Patel S, Ward JM, Woodley JM, Doig SD (2001)
cell density fed-batch cultures in Escherichia coli. Biotechnol Large scale production of cyclohexanone monooxygenase from
Bioeng 41:781–790 Escherichia coli TOP10 pQR239. Enzyme Microb Technol
23. Eiteman MA, Altman E (2006) Overcoming acetate in Escheri- 28:265–274
chia coli recombinant protein fermentations. Trends Biotechnol 40. Xu B, Jahic M, Blomsten G, Enfors S-O (1999) Glucose overflow
24:530–536 metabolism and mixed-acid fermentation in aerobic large-scale
24. Menzella HG, Ceccarelli EA, Gramajo HC (2003) Novel Escheri- fed-batch processes with Escherichia coli. Appl Microbiol Bio-
chia coli strain allows efficient recombinant protein production technol 51:564–571
using lactose as inducer. Biotechnol Bioeng 82:809–817 41. Bylund F, Collet E, Enfors S-O, Larsson G (1998) Substrate gradi-
25. Bruschi M, Krömer JO, Steen JA, Nielsen LK (2014) Production ent formation in the large-scale bioreactor lowers cell yield and
of the short peptide surfactant DAMP4 from glucose or sucrose in increases by-product formation. Bioprocess Eng 18:171–180
high cell density cultures of Escherichia coli BL21 (DE3). Micro- 42. Enfors S-O, Jahic M, Rozkov A, Xu B, Hecker M, Jürgen B,
bial Cell Factories 13:99 Krüger E, Schweder T, Hamer G, O’beirne D (2001) Physiologi-
26. Ramalingam S, Gautam P, Mukherjee KJ, Jayaraman G (2007) cal responses to mixing in large scale bioreactors. J Biotechnol
Effects of post-induction feed strategies on secretory production 85:175–185
of recombinant streptokinase in Escherichia coli. Biochem Eng J 43. Middelberg AP, O’Neill BK, Bogle L, David I, Snoswell MA
33:34–41 (1991) A novel technique for the measurement of disruption in
27. Wong HH, Kim YC, Lee SY, Chang HN (1998) Effect of post- high-pressure homogenization: Studies on E. coli containing
induction nutrient feeding strategies on the production of bioad- recombinant inclusion bodies. Biotechnol Bioeng 38:363–370
hesive protein in Escherichia coli. Biotechnol Bioeng 60:271–276 44. Sauer T, Robinson CW, Glick BR (1989) Disruption of native
28. Gao H-J, Wu Q, Chen G-Q (2002) Enhanced production of D-(– and recombinant Escherichia coli in a high-pressure homogenizer.
)-3-hydroxybutyric acid by recombinant Escherichia coli. FEMS Biotechnol Bioeng 33:1330–1342
Microbiol Lett 213:59–65 45. Matsui I, Sakai Y, Matsui E, Kikuchi H, Kawarabayasi Y, Honda
29. Jensen EB, Carlsen S (1990) Production of recombinant human K (2000) Novel substrate specificity of a membrane-bound
growth hormone in Escherichia coli: expression of different pre- β-glycosidase from the hyperthermophilic archaeon Pyrococcus
cursors and physiological effects of glucose, acetate, and salts. horikoshii. FEBS letters 467:195–200
Biotechnol Bioeng 36:1–11 46. Watson J, Cumming R, Street G, Tuffnell J (1987) Release of
30. Strandberg L, Enfors S-O (1991) Batch and fed batch cultivations intracellular protein by thermolysis, pp 105–109
for the temperature induced production of a recombinant protein 47. Zhao CX, Dwyer MD, Yu AL, Wu Y, Fang S, Middelberg AP
in Escherichia coli. Biotech Lett 13:609–614 (2015) A simple and low-cost platform technology for producing
31. De Anda R, Lara AR, Hernández V, Hernández-Montalvo V, pexiganan antimicrobial peptide in E. coli. Biotechnol Bioeng
Gosset G, Bolívar F, Ramírez OT (2006) Replacement of the glu- 112:957–964
cose phosphotransferase transport system by galactose permease 48. Potvin G, Ahmad A, Zhang Z (2012) Bioprocess engineering
reduces acetate accumulation and improves process performance aspects of heterologous protein production in Pichia pastoris: a
of Escherichia coli for recombinant protein production without review. Biochem Eng J 64:91–105
impairment of growth rate. Metabol Eng 8:281–290 49. Klein-Marcuschamer D, Simmons BA, Blanch HW (2011)
32. Hellmuth K, Korz D, Sanders E, Deckwer W-D (1994) Effect Techno-economic analysis of a lignocellulosic ethanol biorefin-
of growth rate on stability and gene expression of recombinant ery with ionic liquid pre-treatment. Biofuels Bioprod Biorefin
plasmids during continuous and high cell density cultivation of 5:562–569
Escherichia coli TG1. J Biotechnol 32:289–298 50. Janet L (2016) ENERGÍAS RENOVABLES REPORTE DE LA
33. Jin D, Chung BH, Do Hwang Y, Park YH (1992) Glucose-limited SITUACIÓN MUNDIAL, REN21 Paris. France
fed-batch culture of Escherichia coli for production of recombi-
nant human interleukin-2 with the DO-stat method. J Ferment
Bioeng 74:196–198

13