Process development of eicosapentaenoic acid production

Biochemical Engineering Journal 82 (2014) 53–62

Contents lists available at ScienceDirect

Biochemical Engineering Journal journal homepage: www.elsevier.com/locate/bej

Regular article

Process development of eicosapentaenoic acid production Ahmed Abd El razak a,b , Alan C. Ward c , Jarka Glassey a,∗ a

School of Chemical Engineering and Advanced Materials, Newcastle University, UK Botany Department, Faculty of Science, Mansoura University, Egypt c School of Biology, Newcastle University, UK b

a r t i c l e

i n f o

Article history: Received 27 November 2012 Received in revised form 18 October 2013 Accepted 29 October 2013 Available online 6 November 2013 Keywords: Polyunsaturated fatty acids (PUFAs) Eicosapentaenoic acid (EPA) Plackett–Burman (PB) Central composite design (CCD) Bioreactor

a b s t r a c t Eicosapentaenoic acid (EPA), a well-known member of omega-3 fatty acids, is considered to have a significant health promoting role in the human body. It is an essential fatty acid as the human body lacks the ability to produce it in vivo and must be supplemented through diet. Microbial EPA represents a potential commercial source. GC/MS analyses confirmed that bacterial isolate 717, similar to Shewanella pacifica on the basis of 16S rRNA sequencing, is a potential high EPA producer. Two types of bioreactors, a Stirred Tank Reactor (STR) and an Oscillatory Baffled Reactor (OBR), were investigated in order to choose the optimum system for EPA production. The EPA production media was optimised through the selection of media components in a Plackett–Burman (PB) design of experiment followed by a Central Composite Design (CCD) to optimise the concentration of medium components identified as significant in the Plackett–Burman experiment. The growth conditions for the bioreactor, using artificial sea water (ASW) medium, were optimised by applying Response Surface Methodology (RSM). This optimisation strategy resulted in an increase in EPA from 33 mg/l (10 mg/g biomass), representing 8% of the total fatty acids at shake flask level, to 350 mg/l (46 mg/g biomass) representing 25% of the total fatty acids at bioreactor level. During this study the main effects and the interactions between the bioreactor growth conditions were revealed and a polynomial model of EPA production was generated. Chemostat experiments were performed to test the effect of growth rate and temperature on EPA production. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Eicosapentaenoic acid (EPA) is a polyunsaturated fatty acid (PUFA) with 20 carbon atoms and five double bonds. As the first double bond appears on the third carbon atom from the methyl end it is an omega-3 fatty acid. Industrial production of EPA has gained more attention recently due to the proven clinical importance of EPA in reducing the risk of cardiovascular diseases, lowering of plasma cholesterol and decreasing the incidence of breast, colon and pancreatic cancers, in addition it plays an important role in controlling various biological processes as it is a precursor for a number of vital eicosanoid signalling compounds [1,2]. As humans lack the ability to produce EPA in vivo, the main EPA source is through dietary supplements [3]. Although fish oil is the main commercial source of EPA as a dietary supplement, there are many limitations on its wider usage such as high purification cost, complex composition, potential heavy metal contamination and unacceptable odour. Recently, fish oil was found to interfere

∗ Corresponding author at: CEAM, Merz Court, Newcastle University Newcastle upon Tyne NE1 7RU, UK. Tel.: +44 0191 222 7275; fax: +44 0191 222 5292. E-mail address: [email protected] (J. Glassey). 1369-703X/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.bej.2013.10.022

with chemotherapy causing cancer cells to become less sensitive to such treatments [4], due to the presence of 12-oxo-5, 8, 10heptadecatrienoic acid and hexadeca-4, 7, 10, 13-tetraenoic acid which, even in minute quantities, induces resistance to a wide spectrum of chemotherapeutic agents. As a result, microbial EPA may be a promising alternative [5,6]. EPA plays a critical role in the bacterial membrane, especially at low temperatures, as it maintains fluidity of the membrane in extreme cold environments [7], and is essential for cell division and membrane organisation [8]. In addition to its role at low temperature, EPA also plays a role as an antioxidant by protecting the cell from oxygen free radicals [9] and in facilitating the transport of hydrophilic and hydrophobic compounds across the bacterial membrane [10]. PUFA production was studied, one factor at a time (OFAT), in Shewanella sp. GA-22 by [11] demonstrating that it is carbontemperature dependant. Different carbon sources (crude oil, gasoline, glucose, glycerol pyruvate n-tetradecane and Tween) were used as sole carbon source in the media and showed marked influences on PUFA production. In addition, temperature also showed a significant effect on PUFA production with at least a twofold increase from 2% (w/v) at 20 ◦ C to 5% of the total fatty acids at 4 ◦ C. Corn steep liquor and marine industrial waste liquid were

54

A.A. El razak et al. / Biochemical Engineering Journal 82 (2014) 53–62

used as a carbon source for a marine bacteria identified as Shewanella putrefaciens, to reduce the overall fermentation costs [12], achieving 200 mg of EPA per litre of broth. The key controlling factor in PUFA synthesis is the temperature. Temperature, as the sole tested variable, significantly affected PUFA production in bacteria with maximum productivity achieved between 5 and 25 ◦ C with no production reported at temperatures higher than 25 ◦ C [13,14]. The same effect of the temperature was observed, in an OFAT approach, in Shewanella olleyana, decreasing the temperature from 24 ◦ C to 4 ◦ C doubled the production of EPA from 10.2% to 23.6% of the total fatty acids [15]). The main aim of this work was to develop a bioprocess for industrial production of EPA through two approaches: a) defining the composition of an optimal production medium, and b) optimising the growth conditions at a bioreactor scale. A statistical design of experiments was undertaken to achieve this target using a Plackett–Burman (PB) design as an initial screening tool followed by a Central Composite Design (CCD) for optimisation. The statistical designs enabled the assessment of the statistical significance of the main effect of each factor and the interactions among them as well as the development of a predictive model for numerical optimisation of the process conditions. Given the reported antioxidant properties of EPA, the effective dissolved oxygen distribution throughout the bioreactor is potentially an important factor in large scale EPA production. Two bioreactor types with different mass transfer abilities were investigated: the stirred tank reactor (STR) is traditionally used in bioprocessing, whilst the oscillatory baffled reactor (OBR) was reported to provide a higher oxygen transfer rate with kL a values reaching an average of 75% above those achieved in STR for yeast fermentation processes [16]. 2. Materials and methods 2.1. Cultivation conditions and strain identification Five different deep sea core sediment and fluff samples were collected from the Mid Atlantic ridge by research personnel at the Dove Marine laboratory, Newcastle University and kindly provided for this research. After initial screening for EPA production, isolate 717 was identified as the highest EPA producer and was selected for optimisation. A loopful of cell biomass of this isolate, incubated on Bacto Marine Agar (DIFCO 2216) at 20 ◦ C for 48 h, was transferred to a 250 ml sterile flask containing 50 ml of marine broth and incubated at 20 ◦ C in an orbital shaking incubator at 160 rpm for two days. The culture was collected into a 50 ml sterile Falcon tube and centrifuged at 6000 rpm at 4 ◦ C for 15 min. The cell pellets were then transferred into 2 ml sterile centrifuge tubes with 30% glycerol and stored at −20 ◦ C for subsequent use. Genomic DNA was extracted and 16S rRNA was amplified using primers 27f and 1592r [17] by PCR, purified and sequenced using standard methods [18]. The resultant almost complete sequence was blasted against the Genbank database to identify the closest type strains and aligned against sequences in the genus Shewanella retrieved from the GenBank and RDP databases using Clustal W in MEGA 3.1 [19]. The aligned sequences were used to construct a phylogenetic tree based upon Jukes and Cantor distances and the neighbour joining algorithm in MEGA 3.1 [19]. 2.2. Seed culture in artificial sea water A loopful of biomass from culture plates was transferred into a 250 ml flask containing 50 ml of artificial sea water (peptone 3.5 g/l; yeast extract 3.5 g/l; NaCl 23 g/l; MgCl2 5.08 g/l; MgSO4 6.16 g/l; Fe2 (SO4 )3 0.03 g/l; CaCl2 1.47 g/l; KCl 0.75 g/l; Na2 HPO4 0.89 g/l; NH4 Cl

5.0 g/l) [20] and grown at 20 ◦ C in an orbital shaker incubator at 160 rpm for 24 h. 2.3. Growth in production media Growth was performed in 250 ml sterile flasks with 50 ml of medium at the given temperature for two days in an orbital shaker incubator at 160 rpm. The basal medium used was 1 g/l yeast extract, 10 g/l NaCl, 6 g/l MgSO4 and 0.75 g/l KCl, while for the CCD experiments the same basal media excluding yeast extract was used. Final biomass from each flask was collected into a 50 ml Falcon tube and centrifuged at 6000 rpm for 15 min at 4 ◦ C. The cell pellets were transferred into a 1.5 ml screw tube and freeze-dried overnight. 2.4. Fatty acid methyl ester (FAME) preparation 20 mg of freeze dried cells were suspended in 2 ml of 5% methanolic HCl and heated at 70 ◦ C for 2 h in sealed tubes. Fatty acid methyl esters were extracted from the cells with 0.6 ml hexane and then dried under nitrogen gas [21]. 2.5. FAME profiling The single point internal standard method was used for the determination of EPA concentration. Methyl nonadecanoate (≥99.5% GC capillary purity, Sigma–Fluka) was used as an internal standard. Gas chromatography (GC) with flame ionisation detector (FID) on a Hewlett-Packard 5890 series 2 chromatograph, with a SGE forte-BPX70 column; 30 m × 0.25 ␮m film thickness from SGE Analytical Science LTD with helium as a carrier gas was used for FAME profiling .The GC temperature was held at 210 ◦ C for 30 min. GC–MS analysis was performed on an Agilent 7890A GC in split mode, injector at 280 ◦ C linked to an Agilent 5975C MSD with electron voltage 70 eV, source temperature 230 ◦ C, quad temperature 150 ◦ C, and multiplier voltage 1800 V, interface temperature 310 ◦ C, controlled by HP Compaq computer using Chemstation software. The sample (1 ␮l) in hexane was injected using HP7683B auto sampler with the split open. Separation was performed on an Agilent fused silica capillary column (30 m × 0.25 mm i.d.) coated with 0.25 ␮m dimethyl poly-siloxane (HP-5) phase. The GC was temperature programmed from 30 to 130 ◦ C at 5 ◦ C/min then to 300 ◦ C at 20 ◦ C/min and held at the final temperature for 5 min with Helium as the carrier gas (flow rate of 1 ml/min, initial pressure of 50 kPa, split at 10 ml/min). 2.6. Bioreactor cultivations Two bioreactor types were investigated in order to determine the optimum for EPA production. The Stirred Tank Reactor (STR) was an Applikon Biotechnology autoclavable 2 L Rushton turbine bioreactor with power number Po 6, impeller diameter Di 0.045 m and reactor vessel diameter DR 0.105 m. The Oscillatory Baffled Reactor (OBR) was a custom made tall cylindrical glass column, 0.024 m in diameter and 1 m in length with a total volume of 1 L. Orifice plate baffles, 0.001 m thick each, connected by stainless steel rods and arranged periodically 0.036 m apart were inserted into the entire length of the column. The incubation temperature was 20 ◦ C for both types of cultivation. The comparison between the two bioreactor types was carried out by introducing the same power densities (250 W/m3 ) in both reactors for the same process and comparing their performance in terms of biomass and product concentrations. The power densities were calculated as follows:

A.A. El razak et al. / Biochemical Engineering Journal 82 (2014) 53–62

For STR [22]: P/V (W/m3 ) =

Po N 3 Di5

(1)

DR2 h/4

where  is the density of the fluid (kg/m3 ), N is the speed of the stirrer (rps), Di is the diameter of the impeller (m),  is the is the viscosity of the fluid (m2 /s1 ), Po is the power number, DR is the diameter of the reactor vessel (m), and h is the height of the vessel which is occupied by the liquid (m). For OBR [23]: P/V (W/m3 ) =

2Nb 1 − ˛2 3 3 xo ω 3CD2 ˛2

(2) (kg/m3 ),

where  is the density of the fluid Nb is the number of baffles per unit length (m−1 ), ˛ the ratio of the effective baffle orifice area to the tube area, xo is the oscillation amplitude (m), ω the angular oscillation frequency (rad/s1 ) and CD the orifice discharge coefficient. 2.7. Chemostat A continuous stirred-tank reactor (CSTR) was set up to test the effect of different growth rates on EPA production in ASW medium described in section 2.2. Three different growth rates were tested (0.0437, 0.1093 and 0.1530 h−1 representing 20%, 50% and 70% of the maximum specific growth rate max respectively). The effect of six different temperatures (5, 10, 15, 20, 25 and 30 ◦ C) on EPA production was tested at a constant growth rate (50% of the overall max ). Three samples were collected at each temperature for biomass and EPA analysis. The samples were collected from within the steady state which was achieved after three volume changes of media at a given dilution rate. The total length of the cultivation was 350 h and the temperature switches were randomised to avoid any potential systematic errors.

55

where E(Xi) is the effect of the tested variable and Yi+ and Yi− are the calculated responses. The significance level (P-value) of each variable was determined using the Student’s t test (Eq. (4)): T(Xi) =

E(Xi)

(4)

SE

where SE, the standard error of variables, is calculated as the square root of the variance of an effect. Any variable with (P < 0.1) was considered to be significant at 90% level of confidence. The relationship between the response and the significant variables within Plackett–Burman was expressed as a first order polynomial (Eq. (5)), Y = ˇ0 +



ˇi Xi

(i = 1, 2, . . ., k)

(5)

where Y is the calculated response, ˇ0 is model intercept, ˇi is the regression coefficient for each corresponding variable, Xi is the corresponding variable and k is the number of variables.

2.8.2. Optimising central composite design After determining the most significant variables via Plackett–Burman design, a central composite design (CCD) was used to estimate the optimum level of each variable. The CCD matrix included 5-levels for each variable, 6-centre points and star points to estimate the curvature. 20 trials with 3 factors and 5 levels CCD were performed to optimise the potential production medium (Table 4). The basal medium used was, 10 g/l NaCl, 6 g/l MgSO4 and 0.75 g/l KCl. A second CCD was used to optimise the physico-chemical cultivation conditions. The main aim was to study the effect of pH, temperature and DO on the EPA production. The CCD is able to capture the main effect of each factor in addition to the interactions between them using a second order polynomial model (Eq. (6))



The strategy for the optimisation of medium composition was based on the investigation of the main effects of 11 different medium components via a Plackett–Burman design of experiments. The most significant factors were subsequently optimised via a Central Composite Design (CCD) to determine the optimum concentrations for the production medium. The effects of pH, temperature and dissolved oxygen (DO) were tested via a separate CCD design to determine the main effects and interactions between them, on EPA production. These experiments were performed in the STR bioreactor batch cultivations to facilitate the control of the investigated environmental factors.

where ˇ0 is model intercept, ˇi is the regression coefficient for each factor, ˇii is the regression coefficient for square effect and ˇij is the regression coefficient for interaction. Analysis of variance (ANOVA) was carried out using Design Expert 8.0 statistical package (StatEase, Inc, Minneapolis, MN, USA).

E(Xi) =

2



Yi+ − N



Yi−



(3)

ˇii Xii +



Y = ˇ0 +

2.8.1. Initial screening by Plackett–Burman Ten different medium components, in addition to one dummy variable to evaluate the standard error of the experiments, were investigated. Each variable was investigated at a high (+) and a low (−) level (Table 1), where the low level represented the absence of the investigated variable, except for sodium chloride as the isolate requires the presence of NaCl for growth. The main aim of the screening experiment was to compare the significance and the main effect of each factor on the amount of EPA produced by the isolate under investigation. The main effects of each factor were determined using Eq. (3).

ˇi Xi +



2.8. Statistical experimental design

ˇij Xij

(6)

3. Results and discussion 3.1. Identification of the isolate and its ability to produce EPA Isolate 717 was identified as Shewanella sp.717 on the basis of 16S rRNA sequencing (Genbank accession number: JX203388). The isolate was similar to Shewanella pacificaT (Genbank accession number: AF500075), with 17 base differences out of 1478 sequenced bases (98.9% similarity). This difference is sufficient to suggest isolate 717 may be a new species but is not conclusive. The ability to produce EPA was confirmed by GC and GC/MS, where the mass spectrum of the putative EPA was compared to pure standard (Sigma Aldrich, UK) showing over 98% similarity. Isolate 717 produced 9 mg/g dry weight and 33 mg/l of EPA representing 8% of the total fatty acid content, when grown on Artificial Sea Water. The GC/MS analysis of the 717 FAMEs, indicated the complete absence of 12-oxo-5, 8, 10-heptadecatrienoic acid and hexadeca-4,7,10,13-tetraenoic acid that inhibit cancer treatment by chemotherapy.

56

A.A. El razak et al. / Biochemical Engineering Journal 82 (2014) 53–62

Table 1 Variables to be investigated for EPA production via Plackett–Burman design. Code

Variables

Unit

Minimum level (−)

Maximum level (+)

V1 V2 V3 V4 V5 V6 V7 V8 V9 V10 V11

l-Proline Casein Fish peptone Na2 HPO4 Ammonium nitrate Urea Hy-Soy Meat peptone Lactose NaCl Dummy

g/l g/l g/l g/l g/l g/l g/l g/l g/l g/l –

0 0 0 0 0 0 0 0 0 10 Distilled water

2 2 2 0.89 2 2 2 2 2 30 Bi-distilled water

Table 2 Design matrix and responses of the screening Plackett–Burman experiment. Run

1 2 3 4 5 6 7 8 9 10 11 12

Variables

Responses

V1

V2

V3

V4

V5

V6

V7

V8

V9

V10

V11

EPA yield (mg/g)

EPA conc. (mg/l)

EPA %

− + − + + − − − + + − +

+ − + − − − − − + + + +

+ − + + + − + − + − − −

+ − − + − + + − − + − +

− + + − − + + − + + − −

+ + − + − + − − + − + −

+ + − − + − + − − + + −

− − − − + + + − + − + +

+ + + − + + − − − − − +

− + + + − − + − − − + +

− − + + + + − − − + + −

26.47 1.21 5.13 1.19 3.02 1.55 3.97 2.23 4.63 26.17 2.41 3.14

15.28 2.11 9.11 2.34 13.71 6.21 9.89 2.35 5.97 16.35 7.94 12.41

11.62 6.64 7.91 7.12 11.47 5.53 13.43 5.18 9.78 15.39 14.44 10.44

Table 3 Statistical analysis of Plackett–Burman experiment showing the effect, regression coefficient, T-value and P-value for each variable on the EPA production; the P-values of statistical significant variables were shown in bold with a star. Variables

EPA yield (mg/g) Effect

Coefficient

T-value

P-value

Effect

Coefficient

T-value

P-value

Effect

Coefficient

T-value

P-value

V1 (l-Proline) V2 (Casein) V3 (Fish peptone) V4 (Na2 HPO4 ) V5 (Ammonium nitrate) V6 (Urea) V7 (Hy-Soy) V8 (Meat peptone) V9 (Lactose) V10 (NaCl)

−0.30 9.12 1.28 7.30 0.71 −1.10 7.56 −7.28 −0.01 −7.83

−0.20 4.56 0.64 3.65 0.35 −0.52 3.78 −3.64 −0.01 −3.91

−1.09 24.99 3.51 20.01 1.92 −2.83 20.70 −19.94 −0.03 −21.46

0.472 0.025* 0.177 0.032* 0.305 0.216 0.031* 0.032* 0.983 0.030*

0.35 5.07 1.48 3.55 −0.73 −3.99 4.48 1.43 2.32 −2.67

0.17 2.53 0.74 1.77 −0.36 −1.99 2.24 0.71 1.16 −1.34

0.28 5.82 1.71 4.07 −0.57 −4.58 5.14 1.64 2.67 −3.07

0.828 0.010* 0.186 0.027* 0.668 0.020* 0.014* 0.199 0.076* 0.054*

0.45 3.36 0.62 1.35 −0.26 −1.44 4.51 1.87 −1.95 0.16

0.22 1.68 0.31 0.67 −0.13 −0.72 2.25 0.93 −0.97 0.08

0.93 6.84 1.26 2.74 −0.33 −2.94 9.15 3.80 −3.98 0.21

0.422 0.006* 0.296 0.071* 0.796 0.061* 0.003* 0.032* 0.028* 0.866

EPA conc. (mg/l)

3.2. Initial screening by Plackett–Burman Ten different media components were screened as components of a potential production medium. Table 1 shows these variables in addition to the minimum and maximum level used for the Plackett–Burman experiments. Table 2 shows the design matrix and the responses obtained from each trial, whilst Table 3 provides the results of the statistical analysis of the results. Casein, Hy-soy and Na2 HPO4 were found to have a positive statistically significant effect on the ability of the isolate 717 to produce EPA expressed either as concentration, yield or percentage of total fatty acid, with P-values < 0.1 and positive-sign-effect values (values shown in bold with a star in Table 3). l-proline, fish peptone and ammonium nitrate showed no significant effect on the calculated responses. Low concentration of NaCl was preferred for EPA production. Lactose, urea and meat peptone were found to have a negative impact on at least one of the calculated responses (Table 4). Casein, Hy-soy and Na2 HPO4 were selected as production medium components for optimisation of EPA production. Different

EPA %

concentrations of Na2 HPO4 were previously used to test the effect of phosphate on the ability of Phaeodactylum tricornutum to produce EPA [24]. 3.3. Optimising production medium composition Whilst Plackett–Burman design of experiment only allowed the identification of the main effects of the various tested medium components, their interactions were evaluated using data from the CCD design (Table 5). The calculated P-values for each response reveal that in addition to linear effects, two way interactions and quadratic effects could have a significant effect on the ability to produce EPA. For optimisation, second order polynomial order models were developed, using coded units, to predict the optimum combination for each response (Eqs. (7)–(9)). EPA yield = 19.65 + 3.51A + 3.67B − 0.56C − 0.21AB +0.098AC − 0.37A2 − 0.31B2

(7)

A.A. El razak et al. / Biochemical Engineering Journal 82 (2014) 53–62

57

Table 4 Central composite design of variables (in un-coded units) with EPA productions as responses. Run

Variables

1 2 3 4 5 6 7 8 9 10a 11 12 13 14 15 16 17 18 19 20 a

Responses

Casein (g/l)

Hy-soy (g/l)

Na2 HPO4 (g/l)

EPA yield mg/g

EPA conc. (mg/L)

EPA %

8 12 4 4 4 8 12 8 4 8 8 8 12 8 15 1 12 8 8 8

1 12 4 12 12 8 4 8 4 15 8 8 4 8 8 8 12 8 8 8

4 6 2 2 6 0.5 2 4 6 4 4 4 6 4 4 4 2 4 4 7.5

29.71 18.26 26.32 19.76 27.01 32.37 15.04 34.11 12.46 0 30.03 28.22 32.71 29.91 3.89 18.14 4.45 31.92 31.47 23.12

72.47 65.38 49.03 42.63 74.86 86.74 66.46 72.61 43.62 0 70.12 65.99 88.73 69.22 5.46 44.53 13.11 68.01 70.65 61.21

17.13 16.12 18.71 15.19 17.21 18.07 17.16 15.02 18.06 0 14.21 15.87 15.02 13.56 9.68 17.71 9.11 14.05 14.27 18.57

No growth detected in this trial.

Table 5 ANOVA table for the CCD experiment; the P-values of statistical significant terms were shown in bold with a star. Variables

EPA yield (mg/g)

A – Casein B – Hy-Soy C – Na2 HPO4 AB AC BC A2 B2 C2

EPA conc. (mg/l)

EPA%

Sum of squares

F value

P-value Prob > F

Sum of squares

F value

P-value Prob > F

Sum of squares

F value

P-value Prob > F

224.50 179.12 130.12 151.96 154.79 25.50 602.31 419.07 25.54

5.88 4.69 3.41 3.98 4.05 0.67 15.77 10.97 0.65

0.03* 0.05* 0.09* 0.07* 0.06* 0.43 0.01* 0.01* 0.43

3324.50 1125.13 233.42 1495.09 195.81 446.62 2645.22 1252.45 9.88

10.89 3.69 0.75 4.90 0.63 1.49 8.66 4.10 0.029

0.00* 0.07* 0.41 0.04* 0.44 0.24 0.01* 0.06* 0.86

50.32 0.24 31.29 2.16 0.92 15.24 1.35 46.96 27.51

7.53 0.03 4.68 0.31 0.11 2.28 0.18 7.03 4.12

0.01* 0.85 0.04* 0.59 0.74 0.15 0.68 0.02* 0.06*

EPA conc. = 29.82 + 12.02A + 6.99B − 0.67AB − 0.78A2 − 0.53B2 (8) EPA % = 22.80 − 0.44A + 0.1B − 0.5C + 0.031BC −0.10B2 + 0.012C 2

(9)

3.4. Verification The suggested combinations were prepared and the isolate was cultivated. The actual values of EPA were compared to the predicted ones (Table 6). The optimum medium for EPA production in terms of concentration and yield are nearly the same, with the high amount of EPA either due to an increase in biomass or the total fatty acid content. The composition of the medium required to increase the EPA percentage in the fatty acid content is different (see Table 6) and a further medium composition is required to optimise all three responses simultaneously (Table 7). The model predictions underestimate the EPA responses (Table 7) but are sufficiently accurate to facilitate process optimisation. Isolate 717 gave a three-fold increase in EPA production when grown in production media compared to the ASW media, with 32 mg/g and 75 mg/l of EPA representing 18% of total fatty acids

compared to 10 mg/g and 33 mg/l of EPA representing 8% of the total fatty acid, respectively. 3.5. Comparison of STR and OBR performance Fig. 1 clearly shows that the amount of EPA produced in STR was much higher than the amount produced, under the same growth conditions and with ASW as the cultivation medium, in the OBR. This could be due to the high rate of oxygen mass transfer within the OBR compared to STR. Although a high rate of oxygen transfer may be desirable for some organisms in certain bioprocesses, in case of EPA production from the marine isolate, a relatively low rate of oxygen transfer could be favourable. Even for growth, there is a remarkable retardation of growth for isolate 717 within the OBR due to the long lag phase observed as a result of the high rate of oxygen transfer. 3.6. Optimising the operating conditions A 20 run Central Composite Design was applied, using ASW as a cultivation medium, to test the effect of three different growth conditions in the bioreactor: temperature, pH and DO (Table 8). From the ANOVA analysis (Table 9), temperature was found to be the most significant growth factor affecting EPA production whilst DO also showed significant effect on all the responses, pH was found to be significant in terms of EPA concentration and EPA

58

A.A. El razak et al. / Biochemical Engineering Journal 82 (2014) 53–62

Table 6 Optimum media combinations for each response in addition to the predicted and the actual values for each response variable. Target

Media

Maximum EPA yield (mg/g) Maximum EPA conc. (mg/l) Maximum EPA %

Casein (g/l)

Hy-Soy (g/l)

Na2 HPO4 (g/l)

13.34 12.08 7.56

5.52 5.83 3.35

3.98 3.95 1.20

Predicted values

Actual values

29.70 85.40 20.83

34.44(±3.4) 95.32(±6.1) 19.25(±1)

Table 7 The optimum medium combination for EPA production. Media

Predicted values

Actual values

Casein (g/l)

Hy-Soy (g/l)

Na2 HPO4 (g/l)

EPA yield (mg/g)

EPA conc. (mg/l)

EPA %

EPA yield (mg/g)

EPA conc. (mg/l)

EPA %

10.31

6.25

2.2

26.65

71.72

16.6

32.51(±1.8)

75.45(±5)

18.2(±0.8)

A

4

STR OBR

B 14 12 10 EPA (mg/g)

Dry Weight (g/l)

3

2

8 6 4

1

2 0

0 0

20

40 60 Time (hrs)

80

100

0

20

40 60 Time (hrs)

80

100

C

15

EPA %

10

5

0 0

20

40

60

80

100

Time (hrs) Fig. 1. Comparison of the growth and productivity of isolate 717 grown in OBR and STR with 20 ◦ C as the cultivation temperature with power densities of 250 W/m3 : (A) biomass concentration (g/l), (B) EPA yield (mg/g) and (C) EPA percentage of the total fatty acid.

yield but its effect on EPA percentage was found to be insignificant. The ANOVA table indicates that some of the interaction and quadratic effects have a significant impact, especially on EPA yield and concentration, while in the case of EPA percentage their effect was insignificant.

In terms of the interaction between the pH and DO, the optimum pH range for all the calculated responses was between 7–7.5. While the optimum DO for the EPA concentration and yield was relatively low (10–20%), the optimum DO for EPA percentage of the total fatty acid was relatively high (45–55%).

A.A. El razak et al. / Biochemical Engineering Journal 82 (2014) 53–62

59

Table 8 The matrix and responses for the CCD experiment. Run

1 2 3 4 5 6 7 8a 9 10 11 12 13 14 15 16 17 18 19 20 a

Variables

Responses

Dissolved oxygen (%)

pH

Temperature (◦ C)

EPA yield (mg/g)

EPA conc. (mg/l)

EPA %

20 40 60 40 40 70 40 40 40 20 60 10 60 60 40 20 40 40 40 20

6 7 6 7 7 7 7 7 8.5 8 6 7 8 8 7 6 5.5 7 7 8

15 20 15 20 20 20 20 30 20 15 25 20 25 15 20 25 20 10 20 25

32.35 21.54 9.52 22.03 21.88 15.80 20.29 0 18.28 30.58 8.05 26.24 12.55 8.76 21.92 6.33 0.29 21.73 20.88 14.58

120.58 101.61 31.63 100.00 108.00 57.68 102.00 0 60.73 88.81 17.86 98.56 77.66 25.01 99.57 25.86 0.25 87.57 101.61 44.07

16.47 14.60 18.16 14.00 13.80 14.10 14.50 0 13.29 16.23 10.97 4.03 11.06 14.37 14.22 11.07 3.46 21.41 14.22 10.80

The response was too low to be measured accurately.

Table 9 ANOVA table for the CCD experiment of environmental condition optimisation; the P-values of statistical significant terms were shown in bold with a star. Variables

A – DO B – pH C – Temperature AB AC BC A2 B2 C2

EPA yield (mg/g)

EPA conc. (mg/l)

EPA %

Sum of squares

F value

P-value Prob > F

Sum of squares

F value

P-value Prob > F

Sum of squares

F value

P-value Prob > F

294.14 87.20 300.75 2.58 266.81 41.77 36.88 169.62 125.74

18.14 5.38 18.55 0.16 16.46 2.58 2.27 10.46 7.76

0.002* 0.042* 0.001* 0.698 0.002* 0.139 0.162 0.008* 0.019*

1996.79 1528.98 2756.82 326.78 4309.57 1693.62 166.38 8688.55 4931.28

14.81 11.34 20.44 2.42 31.96 12.56 1.23 64.43 36.57

0.003* 0.007* 0.001* 0.151 0.000* 0.005* 0.292 0.000* 0.000*

57.22 6.78 235.02 5.62 1.81 1.85 32.20 11.62 4.10

3.41 0.40 14.03 0.34 0.11 0.09 1.92 0.69 0.21

0.089* 0.536 0.002* 0.573 0.747 0.764 0.191 0.421 0.656

Table 10 Optimum culture condition combinations for each response in addition to the predicted and the actual values for each response variable. Target

Variable

Maximum EPA yield (mg/g) Maximum EPA conc. (mg/l) Maximum EPA %

Dissolved oxygen (%)

pH

Temperature (◦ C)

12.52 16.67 50.00

7.10 7.13 7.43

14.06 14.44 10.00

Predicted values

Actual values

40.38 139.08 23.75

35.5(±4) 150.8(±7) 22.5(±1)

Table 11 The optimum culture condition combination for EPA production. Variable

Predicted values

Actual values

Dissolved oxygen (%)

pH

Temperature (◦ C)

EPA yield (mg/g)

EPA conc. (mg/l)

EPA %

EPA yield (mg/g)

EPA conc. (mg/l)

EPA %

25.28

7.21

10.00

35.35

120.59

19.26

29.8(±2.2)

137.4(±9.8)

18.22(±0.6)

In terms of the interaction between DO and temperature, low temperature was desirable for all EPA response variables with higher DO levels preferable for high EPA percentage compared to EPA concentration and yield. For pH and temperature, a low temperature (10–15 ◦ C) and a pH range of 7–7.5 was found to be the optimum for all the calculated EPA responses. A lower temperature resulting in higher EPA production is in line with the reported ability of EPA to maintain the fluidity of the membrane in cold conditions [7]. An interesting observation is low DO levels (10–15%) being preferable for high EPA yield and

concentration, compared to the high DO levels (40–50%) required for high EPA percentage, suggesting that low dissolved oxygen is optimum for growth and biomass (and the total fatty acid production) leading to high EPA yield and concentration (mg/l) while at high DO levels the cells grow less well but produce high EPA as a percentage of the total fatty acids. The increase in EPA compared to the total fatty acids when cells are exposed to relatively high DO could be due to either the antioxidant capability of EPA [25] or the activation of oxygen-dependant enzymes in the desaturation and elongation of PUFAs [26] though the production of PUFAs by the polyketide pathway is oxygen independent.

60

A.A. El razak et al. / Biochemical Engineering Journal 82 (2014) 53–62

50

A

350

B

300

40 EPA (mg/l)

EPA (mg/g)

250

30

20

200 150 100

10 50 0

0 ASW

ASW

PM

30

PM Media Type

Media type

C

25

EPA %

20 15 10 5 0 ASW

PM Media Type

Fig. 2. Comparison of the EPA production in ASW medium and production medium (PM), for isolate 717, under the same growth conditions in the bioreactor (conditions summarised in Table 10).

These results support the observation that STR resulted in higher EPA production than OBR (Fig. 1), as high oxygen levels showed a negative impact on EPA production especially in terms of the yield and concentration. For optimisation, second order polynomial order models were developed, using coded units, to predict the optimum combination for each response (Eqs. (10)–(12)). EPA yield = 21.02 − 3.92A + 2.74B − 4.45C + 5.62AC −4.50B2 − 2.45C 2

The optimum conditions for EPA production in terms of concentration and yield are nearly the same, where the high amount of EPA is due to the increase in biomass and total fatty acid content. The medium required to increase the EPA percentage is different and so is the medium required to reach the optimum value for all three EPA response variables (Table 11). Overall process optimisation would depend upon determining the weightings applied to each variable in a cost model for both production and downstream processing.

(10) 3.8. Comparison of EPA production in ASW and production Media

EPA conc. = 94.41 − 13.73A + 10.43B − 13.47C + 22.59AC +14.55BC − 28.98B2 − 13.34C 2 EPA % = 12.57 + 1.36A − 4.01C

(11)

(12)

3.7. Verifying the model These models were verified experimentally and predicted responses were compared against the actual values (Table 10).

Under the optimum bioreactor cultivation conditions for EPA production by isolate 717, a comparison was performed between the ASW and the optimised production media (PM) (Fig. 2). The difference between the ASW and the PM in the ability to produce EPA either as concentration or yield is significant (Fig. 2). The amount of EPA (mg/g) produced when growing on the production medium was approximately 46 mg/g while on the ASW it was 32 mg/g. The amount of EPA produced when growing on the production medium was approximately 350 mg/l which is more than double of that produced when growing on the ASW medium

A.A. El razak et al. / Biochemical Engineering Journal 82 (2014) 53–62 Dry Weight (mg/l) EPA % EPA (mg/g)

3.5 25 3.0

Dry Weight

2.0

15

1.5

10

1.0

EPA

20

2.5

5

0.5 0 0.0 5

10

15

20

25

30

Temperature Fig. 3. EPA production by the 717 isolate in a chemostat under various temperatures at a constant growth rate of 0.1093 h−1 representing 50% of the maximum specific growth rate max .

(150 mg/l). The difference in terms of EPA% of total fatty acids were an increase from 8% in the shake flask to 23% in bioreactor. For production medium, in other experiments, the EPA% increased from 19% in shake flasks to 25% in the bioreactor indicating that 25% could be possible. The amount of EPA produced by isolate 717 in the bioreactor was significantly increased compared to the amount of EPA produced in shake flasks in all calculated responses, either in ASW or the production media. 3.9. Chemostat In the first set of chemostat experiments, the effect of specific growth rates was assessed and it was found to have a negligible effect on EPA production. As temperature was found to be the most significant factor affecting EPA production, another set of chemostat experiments were performed to test that effect at constant growth rate (Fig. 3). The optimum temperature for growth and EPA yield was found to be 15 ◦ C. For EPA percentage an inversely proportional relationship was found between temperature and EPA percentage as EPA decreases with increasing the temperature from the highest percentage at 5 ◦ C to very low levels at 30 ◦ C. The results show that lower temperatures are desirable for higher EPA production particularly in term of the EPA percentage. This result supports the theory that EPA plays a role in the cold adaptation mechanism by sustaining the membrane fluidity at low temperature [7,27]. 4. Conclusions Bacteria can be considered as a substantial resource for the production of PUFA. Bacterial PUFAs could be used as a dietary supplement or introduced into the food chain by using them as a food stock for fish in fish farms. During this research the effect of the cultivation conditions on the ability to produce EPA within the bioreactor were investigated for the first time using a statistical design of experiments, testing the effect of temperature, dissolved oxygen and pH and revealing the significance of their effect, interactions between them and the quadratic effect of each variable. Optimising the cultivation conditions in the bioreactor resulted in a significant increase in the amount of EPA produced and could be

61

considered as a first step towards economically viable production of EPA. The statistical design of experiment and response surface methodology used in this research proved to be a successful optimisation strategy for developing a production medium and determining the optimum cultivation conditions. With 46 mg/g dry weight (350 mg/l production medium) and 25% of the total fatty acids, the isolate 717 can be considered as one of the highest reported bacterial EPA producers. It can produce more than twice the amount of EPA reported for Phaeodactylum tricornutum, which produced 133 mg/l of culture media [24], and many fold more than the amount produced by Shewanella marinintestina strain IK-1 which is able to produce 8 mg/l after the addition of cerulenin to its production medium [28]. In addition in the total fatty acids produced by isolate 717 the EPA percentage is at least twice that reported for Shewanella sp. KMG427 and Photobacterium sp. SAMA2 which produced 10% and 13.9% EPA, respectively [29]. This was achieved by optimisation without genetic manipulation, thus avoiding any ethical concerns about the use of genetically modified organisms in the production of nutraceuticals for human consumption.

References [1] J. Frank, F.J. Kelly, The metabolic role of n–3 polyunsaturated fatty acids: relationship to human disease, Comparative Biochemistry and Physiology Part A: Physiology 98 (3–4) (1991) 581–585. [2] R. De Caterina, n–3 fatty acids in cardiovascular disease, New England Journal of Medicine 364 (25) (2011) 2439–2450. [3] I. Gill, R. Valivety, Polyunsaturated fatty acids. Part 1: Occurrence, biological activities and applications, Trends in Biotechnology 15 (10) (1997) 401–409. [4] J.M.L. Roodhart, L.G.M. Daenen, E.C.A. Stigter, H.J. Prins, J. Gerrits, J.M. Houthuijzen, M.G. Gerritsen, H.S. Schipper, M.J.G. Backer, M. Van Amersfoort, J.S.P. Vermaat, P. Moerer, K. Ishihara, E. Kalkhoven, J.H. Beijnen, P.W.B. Derksen, R.H. Medema, A.C. Martens, A.B. Brenkman, E.E. Voest, Mesenchymal stem cells induce resistance to chemotherapy through the release of platinum-induced fatty acids, Cancer Cell 20 (3) (2011) 370–383. [5] Z.Y. Wen, F. Chen, Heterotrophic production of eicosapentaenoic acid by microalgae, Biotechnology Advances 21 (4) (2003) 273–294. [6] R. Colin, Fatty acid biosynthesis in microorganisms being used for single cell oil production, Biochimie 86 (11) (2004) 807–815. [7] S. Sato, T. Kurihara, J. Kawamoto, M. Hosokawa, S. Sato, N. Esaki, Cold adaptation of eicosapentaenoic acid-less mutant of Shewanella livingstonensis Ac10 involving uptake and remodeling of synthetic phospholipids containing various polyunsaturated fatty acids, Extremophiles 12 (6) (2008) 753–761. [8] J. Kawamoto, T. Kurihara, K. Yamamoto, M. Nagayasu, Y. Tani, H. Mihara, M. Hosokawa, T. Baba, S.B. Sato, N. Esaki, Eicosapentaenoic acid plays a beneficial role in membrane organization and cell division of a cold-adapted bacterium, Shewanella livingstonensis Ac10, Journal of Bacteriology 191 (2) (2009) 632–640. [9] H. Okuyama, Y. Orikasa, T. Nishida, Significance of antioxidative functions of eicosapentaenoic and docosahexaenoic acids in marine microorganisms, Applied and Environmental Microbiology 74 (3) (2008) 570–574. [10] T. Nishida, R. Hori, N. Morita, H. Okuyama, Membrane eicosapentaenoic acid is involved in the hydrophobicity of bacterial cells and affects the entry of hydrophilic and hydrophobic compounds, FEMS Microbiology Letters 306 (2) (2010) 91–96. [11] G. Gentile, V. Bonasera, C. Amico, L. Giuliano, M.M. Yakimov, Shewanella sp. GA-22, a psychrophilic hydrocarbonoclastic antarctic bacterium producing polyunsaturated fatty acids, Journal of Applied Microbiology 95 (2003) 1124–1133. [12] K. Yazawa, Production of eicosapentaenoic acid from marine bacteria, Lipids 31 (1) (1996) S297–S300. [13] D.S. Nichols, T.A. McMeekin, Biomarker techniques to screen for bacteria that produce polyunsaturated fatty acids, Journal of Microbiological Methods 48 (2-3) (2002) 161–170. [14] S.H. Yang, J.H. Lee, J.S. Ryu, C. Kato, S.J. Kim, Shewanella donghaensis sp. nov., a psychrophilic, piezosensitive bacterium producing high levels of polyunsaturated fatty acid, isolated from deep-sea sediments, International Journal of Systematic and Evolutionary Microbiology 57 (2) (2007) 208–212. [15] J.H. Skerratt, J.P. Bowman, P.D. Nichols, Shewanella olleyana sp. nov., a marine species isolated from a temperate estuary which produces high levels of polyunsaturated fatty acids, International Journal of Systematic and Evolutionary Microbiology 52 (6) (2002) 2101–2106. [16] X. Ni, S. Gao, R.H. Cumming, D.W. Pritchard, A comparative study of mass transfer in yeast for a batch pulsed baffled bioreactor and a stirred tank fermenter, Chemical Engineering Science 50 (13) (1995) 2127–2136.

62

A.A. El razak et al. / Biochemical Engineering Journal 82 (2014) 53–62

[17] D.J. Lane, 16S/23S rRNA sequencing, in: E. Stackebrandt, M. Goodfellow (Eds.), Nucleic Acid Techniques in Bacterial Systematics, John Wiley & Sons, Chichester, United Kingdom, 1991, pp. 115–175. [18] J. Sambrook, D.W. Russell, Molecular Cloning: A Laboratory Manual, 3rd ed., Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, 2001. [19] S. Kumar, K. Tamura, M. Nei, MEGA3: integrated software for molecular evolutionary genetics analysis and sequence alignment, Briefings in Bioinformatics 5 (2004) 150–163. [20] S. Lang, M. Hüners, V. Lurtz, Bioprocess engineering data on the cultivation of marine prokaryotes and fungi marine biotechnology II, in: R. Ulber, G.Y. Le (Eds.), Advances in Biochemical Engineering/Biotechnology, vol. 97, Springer, Berlin/Heidelberg, 2005, p. 585. [21] K. Watanabe, C. Ishikawa, K. Yazawa, K. Kondo, A. Kawaguchi, Fatty acid and lipid composition of an eicosapentaenoic acid-producing marine bacterium, Journal of Marine Biotechnology 4 (1996) 104–112. [22] F.A. Holland, F.S. Chapman, Liquid Mixing and Processing in Stirred Tanks, Reinhold Publishing Corporation, USA, 1966. [23] M.R. Hewgill, M.R. Mackley, A.B. Pandit, S.S. Pannu, Enhancement of gas–liquid mass transfer using oscillatory flow in a baffled tube, Chemical Engineering Science 48 (4) (1993) 799–809.

[24] W. Yongmanitchai, O.P. Ward, Growth of and omega-3 fatty acid production by Phaeodactylum tricornutum under different culture conditions, Journal of Applied and Environmental Microbiology 57 (2) (1991) 419–425. [25] T. Nishida, R. Hori, N. Morita, H. Okuyama, Membrane eicosapentaenoic acid is involved in the hydrophobicity of bacterial cells and affects the entry of hydrophilic and hydrophobic compounds, FEMS Microbiology Letters 306 (2) (2010) 91–96. [26] K. Higashiyama, K. Murakami, H. Tsujimura, N. Matsumoto, S. Fujikawa, Effects of dissolved oxygen on the morphology of an arachidonic acid production by Mortierella alpina 1S-4, Biotechnology and Bioengineering 63 (4) (1999) 442–448. [27] E.F. Delong, A.A. Yayanos, Biochemical function and ecological significance of novel bacterial lipids in deep-sea prokaryotes, Journal of Applied and Environmental Microbiology 51 (4) (1986) 730–737. [28] M. Naoki, T.N. Mika, Y. Yutaka, O. Hidetoshi, Enhancement of polyunsaturated fatty acid production by cerulenin treatment in polyunsaturated fatty acidproducing bacteria, Biotechnology Letters 27 (6) (2005) 389–393. [29] W.H. Lee, K.W. Cho, S.Y. Park, K.S. Shin, D.S. Lee, S.K. Hwang, S.J. Seo, J.M. Kim, S.Y. Ghim, B.H. Song, S.H. Lee, J.G. Kim, Identification of psychrophile Shewanella sp KMG427 as an eicosapentaenoic acid producer, Journal of Microbiology and Biotechnology 18 (12) (2008) 1869–1873.