Bioseparation 7: 915, 1997. c 1997 Kluwer Academic Publishers. Printed in the Netherlands.

Pilot-scale extraction of PHB from recombinant E. coli by homogenization and centrifugation

Y. Ling1, H.H. Wong1 , C.J. Thomas2, D.R.G. Williams1 & A.P.J. Middelberg1
1

Department of Chemical Engineering and 2 Department of Microbiology and Immunology, University of Adelaide, S.A. 5005, Australia

Received 29 May 1997; accepted in revised form 19 November 1997

Key words: centrifugation, extraction, homogenization, poly- -hydroxybutyrate (PHB)

Abstract A new method of poly- -hydroxybutyrate (PHB) extraction from recombinant E. coli is proposed, using homogenization and centrifugation coupled with sodium hypochlorite treatment. The size of PHB granules and cell debris in homogenates was characterised as a function of the number of homogenization passes. Simulation was used to develop the PHB and cell debris fractionation system, enabling numerical examination of the effects of repeated homogenization and centrifuge-feedrate variation. The simulation provided a good prediction of experimental performance. Sodium hypochlorite treatment was necessary to optimise PHB fractionation. A PHB recovery of 80% at a purity of 96.5% was obtained with the nal optimised process. Protein and DNA contained in the resultant product were negligible. The developed process holds promise for signicantly reducing the recovery cost associated with PHB manufacture. Introduction Poly- -hydroxybutyrate (PHB) is a bacterial storage polymer produced in various microorganisms in response to nutrient limitation. PHB was rst described by Lemoigne in Bacillus megaterium in 1926 (Dawes and Senior, 1973). Its copolymer P(HB-co-HV) has been marketed as a biodegradable thermoplastic since the middle 1980s. The high production cost of PHB and its copolymer has restricted their use as bulk plastic materials. Production economics are governed by several factors including product yield, product recovery, and the techniques and strategy applied. It has been estimated that over half of the production cost of PHB is associated with the recovery and purication process. Moreover, the recovery process can signicantly alter the properties of the nal product. While a number of PHB recovery methods have been developed (Hahn et al., 1994; Berger et al., 1989; Lee et al., 1995), most have not been tested at pilot nor production scale. Industrial methods for PHB release include enzymatic digestion (Holmes and Lim, 1990) and thermal treatment coupled with high-pressure homogenization (Harrison, 1990). A high disruption efciency for A. eutrophus is obtained using high-pressure homogenization, and key factors affecting disruption have been thoroughly investigated (Harrison et al., 1991). However, processes for removing cell debris from PHB following cell disruption have received little attention. The process reported by Harrison (1990) was based on solubilisation of non-PHB components using detergent and enzymes coupled with multiple centrifugation passes. This approach raises total production cost and still yields a relatively low quality PHB, which is unacceptable for many applications. For historic reasons, these extraction processes are mostly focused on A. eutrophus. However, research interest has progressively shifted to recombinant E. coli. The advantages of using recombinant E. coli strains as a host for PHB accumulation have been detailed by Lee and Chang (1996). Effectiveness of PHB recovery varies from one host to another due to differences in cell-wall structure and PHB formation factors. Consequently, differences in cell disruption and the subsequent fractionation process are expected.

10 For example, it has been reported that the accumulation of high levels of PHB in recombinant E. coli facilitates PHB granule release (Lee, 1996). Centrifugation is one of the most common unit operations in bioprocessing for the fractionation of insoluble protein inclusion bodies from other cellular material (Middelberg, 1996). Fractionation efciency can be enhanced by optimising centrifugation conditions (e.g. feedrate), even though perfect fractionation is never achieved. Purity can be enhanced through multiple centrifuge passes, minimising the need for expensive chemicals and enzymes. Chemical treatment for further purication can be considered if a high purity product is required. Despite the widespread use of centrifugation for the recovery of inclusion bodies, this technique has not been thoroughly investigated as a method for collection of PHB granules. Fractionation by centrifugation is dictated by the relative sedimentation-velocity distributions of the granules and cell debris (e.g. size and density) (Hoare and Dunnill, 1989). A rational understanding of PHB and cell debris characteristics is therefore required. The size distribution of PHB granules is easily determined by analytical disc centrifuge (CDS) (Middelberg et al., 1995) and a mean PHB granule diameter of 1.131.25 m has been reported in recombinant E. coli, based on a density of 1260 kg/m3. Cell debris is, however, more difcult to characterise. Various techniques are available, but all have limitations (Wong et al., 1996). Recently, Wong et al. (1996) introduced a new analysis method (Cumulative Sedimentatian Analysis or CSA) for characterising cell-debris size based on cumulative sedimentation under centrifugal force, coupled with analysis of the sedimented fraction by SDSPAGE. This method provides a means to characterise cell debris size and thus optimise centrifugal fractionation. In this paper, CSA is used to characterise cell-debris size distributions following the repeated homogenization of recombinant E. coli containing PHB granules. Based on these experimental data, the fractionation of PHB and cell debris by homogenization and centrifugation is simulated. A scaled up PHB extraction process combining homogenization, disc-stack centrifugation and chemical treatment is nally suggested and tested. The process gives PHB with a high purity and recovery, and is amenable to scale up. Theoretical note Boltzmann equation The Boltzmann function, Equation (1), has been applied to describe yeast and E. coli debris distributions following homogenization (Siddiqi et al., 1996; Wong et al., 1997). The median diameter D50 and parameter w were obtained by regression of experimental data to the equation.

W D  = 

1 + exp

D D50
 w

(1)

Grade efciency curve The particle fractionation by a centrifuge can be evaluated using a grade-efciency curve, which describes the size-dependent particle classication. One of the grade-efciency curves for disc-stack centrifugation, based on a complete radial (lateral) mixing model assumption, is given by Equation (2) (Mannweiler, 1989).

T D  = 1

exp

n  D k D c 

(2)

The critical diameter, Dc , depends upon centrifuge operating conditions and particle properties. Parameters k and n can be obtained by regression to experimental data.

Materials and methods Bacterial strain and fermentation A non-K-12 E. coli strain Topp1 [F, proAB, lacIq , Z M15, Tn10, (tetR )) (Stratagene, CA, USA) was transformed with plasmid pJM9123 (Slater et al., 1992). Transformants were selected using kanamycin (50 g mL 1 ). A fed-batch fermentation using this strain was conducted in modied R-medium, fed intermittently with a mixture of glucose, yeast extract, and MgSO4
7H2 O. Detailed descriptions of the fermentation conditions used are given elsewhere (Ling et al., 1997). The fermentation gave a total of 18 L of culture with 42.3 g L 1 cell concentration (DCW). The culture was stored in the fermenter overnight at 10 C before homogenization for debris size analysis (4 L) or storage at 18 C for subsequent use (14 L).

11

Figure 2. Size distributions for whole cells and PHB granules in homogenates determined by photosedimentation. N is the number of homogenization passes.

Figure 1. PHB recovery protocols.

Extraction process and analytical methods Homogenization and debris size measurement Fresh culture harvested from the fermentation was diluted in PBS buffer (1.37 g L 1 KH2 PO4 , 6.5 g L 1 NaCl, adjusted pH to 6.9 with 2 M NaOH) to give an optical density of 1215 at 600 nm. Five discrete homogenisation passes were conducted at 55 MPa in an APV-Gaulin high-pressure homogenizer (15MR, CD valve). The feed temperature was approximately 10 C for each pass. 1 L of homogenate was collected after each pass. CSA was conducted for the measurement of cell-debris size distributions (Wong et al., 1997). Cell disruption and PHB release by homogenization were monitored by the use of an Applied Imaging DCF4 Disc Centrifuge (Gateshead, UK) with a standard water spin uid scheme (Middelberg et al., 1990). The density of whole cells and the resulting cellular debris was taken as 1085 kg m 3 (Hwang, 1996), and PHBgranule density was taken as 1260 kg m 3 (Middelberg et al., 1995). These densities are approximate and only used to transform distributions to a size basis for presentation. In simulation, the settling-velocity distribution is the critical determinant of performance, and this is not affected by the assumed density. Frozen culture from the fermentation was thawed at 4 C and then diluted ten times with PBS buffer (as above). Figure 1 summarises the three extraction processes and operating conditions employed. Centrifugation was performed with a Veronesi KLE-160 solidbowl disc-stack centrifuge with a xed operating speed P of 8400 rpm ( =3775 m2 ). PHB collection efciency during centrifugation was determined by comparing feed and supernatant samples using an analytical disc centrifuge (CDS). Other analytical methods (e.g. PHB quantitation by GC) were as previously described (Ling et al., 1997). Hypochlorite treatment Sodium hypochlorite (NaOCl) (Fisons technical grade, FSE Australia, 100 g active chlorine L 1 ) was diluted with RO water to give a solution with 57 g active chlorine L 1 (Middelberg et al., 1995). This was added to samples to give 0.85 g active chlorine L 1 corresponding to 1.5% NaOCl relative to the stock solution, which were then incubated for 1 h at a temperature between 15 C to 20 C, prior to centrifugation. The centrifugation took about 40 minutes (Figure 1).

12
Table 1. Median Diameters (D50 , m) and Boltzmann parameters (w) for whole cells, PHB granules, and cell debris in homogenates measured by CSA or CDS (errors represent standard deviation of the mean by regression to Equation (1)) Whole cells D50 w D50 w N=1 N=2 N=3 N=4 N=5 N=1 N=2 N=3 N=4 N=5 1.38 0.139 0.58 0.084 0.65 0.50 0.44 0.38 0.36 0.20 0.17 0.16 0.12 0.10

PHB in homogenates

D50 Cell debris in homogenates w

 0.010  0.016  0.017  0.010  0.010  0.009  0.016  0.018  0.010  0.009

Figure 3. Cumulative oversize distributions of cell debris in homogenates as a function of the number of homogenizer passes (N) determined by cumulative sedimentation analysis. Smooth curves were obtained by regression to equation (1) with parameters given in Table 1.

Results and discussion Preliminary experimental investigation The fermentation gave a cell broth with a dry weight of 42.3 g L 1 and a PHB content of 52% (w/w). Figure 2 presents the particle size distributions of homogenates as a function of the number of homogenizer passes, measured by CDS. PHB mean diameter after the third homogenizer pass was centred at 0.62 m. The PHB granules in this study are smaller than those reported previously (Middelberg et al., 1995), but are larger than recombinant IGF inclusion bodies recovered successfully by centrifugation (Wong et al., 1996). Obviously, high cell disruption was obtained after two homogenizer passes. Further homogenization apparently caused some PHB granule breakage. This has also been observed during the homogenization of frozen cells containing PHB granules (Ling et al., 1997). Cell debris is comminuted with each homogenization pass (Wong et al., 1997). Cell debris size after each homogenizer pass was measured by CSA and regressed to Equation (1). Figure 3 presents the cumulative oversize distributions of cell debris in homogenates as a function of the number of homogenizer passes. Smooth curves are regressions to Equation (1). The size distribution of whole cells measured by CDS is also included for comparison. Table 1 lists the Boltzmann parameters (D50 and w) determined by regression. The Boltz-

mann equation describes the cell-debris distributions for E. coli homogenates containing PHB granules very well. The results conrm that repeated homogenization causes micronisation of cellular debris. Median debris size decreased from 0.65 m after the rst homogenizer pass to 0.36 m after ve homogenizer passes. Median debris size and distribution width after each homogenizer pass are comparable with those for E. coli containing recombinant protein inclusion bodies under similar homogenization conditions (Wong et al., 1997). Simulation study: design of an extraction process Cell debris represents the insoluble contaminant in PHB homogenates and thus is the major concern during fractionation by centrifugation. Based on the information on PHB granule and cell-debris sizes obtained above, the settling characteristics of PHB granules and cell debris in a disc-stack centrifugation can be studied. PHB collection efciency and cell debris removal by homogenization and centrifugation can be predicted by simulation, and a possible fractionation scheme developed. The settling behaviour of cell debris in a disc-stack centrifugation is described by Equation (2), assuming the parameters k and n are 0.13 and 2.1, respectively, obtained for recombinant E. coli cell debris produced by homogenization and fractionated in the same cen-

13 Table 2. Actual cell debris removal was 75% (A1) at a feedrate of 3.09  10 9 m s 1 , which is comparable to the simulated value for 3 homogenizer passes (70%). The small deviation could be due to several factors. Firstly, cell material in the pilot scale tests was frozen and stored before homogenization. Prolonged storage and freeze-thaw cycles can promote changes in cell wall characteristics and integrity, affecting the actual debris size and also homogenate properties such as viscosity. Secondly, an assumed viscosity of 1.2  10 3 Pa.s was used in the simulation. Actual viscosity could be higher, leading to a higher removal of debris. The predicted PHB collection efciency (87%) was signicantly higher than the experimental value (76%). The errors indicated above will affect the prediction. Furthermore, the parameters in Equation (2) used in the simulation were determined for recombinant protein granules with a median diameter of approximately 0.45 m. Extrapolation to the larger PHB granules in this study may introduce errors. Extraction processing trifuge used in this study (Wong, 1996). The predicted effect of repeated homogenization and centrifugefeedrate variation on the removal of cell debris is presented in Figure 4. The simulation P is limited to normalised centrifuge feedrates (Q/ ) between 1.32  10 9 m s 1 to 3.97  10 9 m s 1 and to 1 to 5 homogenizer passes, as this is the range of model and parameter condence. Cell debris removal depends strongly on the centrifuge feedrate and the number of homogenizer passes as expected. Using the same number of homogenizer passes, debris removal increases signicantly as the centrifuge feedrate increases. PHB recovery decreases concomitantly. Clearly, there is a strong interaction between the homogenization and centrifugation unit operations. Figure 4 also includes the curve describing PHB overall collection efciency as a function of centrifuge feedrate, using parameters k and n in Equation (2) of 0.16 and 2.6 (Wong, 1996) and the PHB size distribution measured after three homogenizer passes. Note that the PHB size distribution is relatively insensitive to the number of homogenizer passes as the extent of granule breakage is low (Figure 2). Overall PHB recovery remains high even at high feedrates (Figure 4), possibly due to the large granule size and high density. Based on the simulation, a fractionation process including two centrifugations was tested experimentally (process A, Figure 1). The results are listed in In process A, 72% PHB recovery and 94% debris removal were achieved and a PHB purity of 90.2% (w/w) was obtained by repeated centrifugation (A2). Centrifugation was also efcient at removing protein and DNA. However, the resultant protein and DNA content were not below the limits reported to cause PHB decolouration in thermal processing (Harrison, 1990), leaving scope for further improvement. Overall PHB recovery in process A was also considered too low. To enhance PHB purity, sodium hypochlorite (NaOCl) treatment was introduced (process B, Figure 1). The use of NaOCl at a relatively low concentration effects the degradation of non-PHB cellular materials and has a negligible effect on PHB granules contained in recombinant E. coli cells (Hahn et al., 1954). Cell debris micronisation, probably coupled with a reduction in viscosity, led to enhanced PHB collection efciency and debris removal (B1, Table 2). Protein removal was also improved by NaOCl treatment. However, DNA removal did not follow the same trend, possibly due to DNA denaturation by NaOCl treatment and adsorption to the granule surface (Ling et al., 1997). To overcome the problem of DNA contamination, it was decided to introduce the NaOCl treatment after the rst centrifugal fractionation (i.e. after the removal of considerable DNA) (process C, Figure 1). Additional

Figure 4. Simulated PHB recovery and cell debris removal in a centrifuge. N is the number of homogenization passes.

14
Table 2. Results for PHB fractionation by centrifugation Extraction process PHB (% w/w) Protein (% w/w) DNA (% w/w) Cell debris removal (%) Collection efciency (% per pass)

Cell broth Process A A1 A2 B1 B2 C1 C2 C3

52.0 79.1 90.2 82.1 94.0 70.2 90.8 96.5

25 9.0 2.3 3.9 1.0 9.1 0.54 <0.01

5.2 1.2 0.62 2.2 1.9 1.2 0.15 0.03

0 75.0 94.3 85.4 94.3 60.8 92.4 96.5 76 95 92 97 82 99 98

Process B

Process C

homogenization steps were introduced after centrifugation to enhance paste re-suspension and hence the dispersion of granules and residual debris. This process gave both high PHB purity (96.5%) and recovery (81%) (Table 2). High PHB recovery is attributed to the reduction in feedrate (C1) and the effect of NaOCl treatment on homogenate viscosity (C2). A comparison of C1 and A1 conrms that collection efciency was improved as the feedrate decreased from 3.09  10 9 m s 1 to 2.21  10 9 m s 1 , but with a penalty of decreased cell debris removal (75% to 61%). These values match reasonably well with the simulated values of 70% and 62%, respectively. DNA removal was enhanced in process C as expected. The nal recovery process yielded a very pure PHB product with negligible DNA and protein contamination. This process is quite competitive in terms of PHB purity and recovery with the industrial method described by Harrison (1990), but does not require enzyme, detergent and hydrogen peroxide washing. The remaining protein and DNA contents are below the levels reported to cause PHB decolouration in thermal processing (Harrison, 1990). This fractionation process was conducted at pilot scale, and the results obtained can be simply extrapolated to full-scale PHB manufacture. Fundamental data on debris size was also obtained. Further process simplication will be possible by using a continuous rather than solid bowl centrifuge, thus simplifying paste re-suspension and recentrifugation. Savings are also possible by optimising the NaOCl treatment regime.

References
Berger E, Ramsay BA, Ramsay JA and Chavarie C (1989) PHB recovery by hypochlorite digestion of non-PHB biomass. Biotechnol. Tech. 3(4): 227232. Dawes EA and Senior PJ (1973) The role and regulation of energy reserve polymers in micro-organisms. Adv. Microb. Physiol. 14: 135266. Hahn SK, Chang YK, Kim BS and Chang HN (1994) Optimisation of microbial poly(3-hydroxybutyrate) recovery using dispersion of sodium hypochlorite solution and chloroform. Biotechnol. Bioeng. 44: 256261. Harrison ST (1990) The extraction and purication of poly- hydroxybutyrate from Alcaligenes eutrophus. PhD dissertation University of Cambridge, UK. Harrison STL, Chase HA and Dennis JS (1991) The disruption of Alcaligenes eutrophus by high pressure homogenization: key factors involved in the process. Bioseparation 2: 155166. Hoare M and Dunnill P (1989) Biochemical engineering challenges of purifying useful proteins. Phil. Trans. R. Soc. Lond. B324: 497501. Holmes PA and Lim GB (1990) Separation Process. U.A. Patent 4910145. Hwang SO (1996) Effect of inclusion bodies on the buoyant density of recombinant Escherichia coli. Biotechnol. Tech. 10(3): 157 160. Lee IY, Chang HN and Park YH (1995) A simple method for recovery of microbial poly- -hydroxybutyrate by alkaline solution treatment. J. Microbiol. Biotechnol. 5(4): 238240. Lee SY (1995) Poly(3-hydroxybutyrate) extrusion by cells of recombinant Escherichia coli. J. Microbiol. Biotechnol. 6(2): 147149. Lee SY and Chang HN (1996) Characteristics of poly(3hydroxybutyric acid) synthesis by recombinant E. coli. Annals of the New York Academy of Science 782: 133142. Ling Y, Williams DRG, Thomas CJ and Middelberg APJ (1997) Recovery of poly-3-hydroxybutyrate from recombinant Escherichia coli by homogenization and centrifugation. Biotechnol. Techn. 11(6): 409412. Mannweiler K (1989) The recovery of biological particles in highspeed continuous centrifuges with special reference to feed-zone break-up effects. PhD dissertation University of London.

15
Middelberg APJ (1996) Large-scale recovery of recombinant protein inclusion bodies expressed in Escherichia coli. J. Microbiol. Biotechnol. 6(4): 225231. Middelberg APJ, Bogle IDL and Snoswell MA (1990) Sizing biological samples by photosedimentation techniques. Biotechnol. Prog. 6: 255261. Middelberg APJ, Lee SY, Martin J, Williams DRG and Chang HN (1995) Size analysis of poly(3-hydroxybutyric acid) granules produced in recombinant Escherichia coli. Biotechnol. Lett. 17(2): 205210. Siddiqi SF, Titchener-Hooker NJ and Shamlou PA (1996) Simulation of particle size distribution changes occurring during highpressure disruption of bakers yeast. Biotechnol. Bioeng. 50: 145150. Slater SC, Gallaher T and Dennis DE (1992) Production of poly-(3-hydroxybutyrate-co-3-hydroxyvalerate) in a recombinant Escherichia coli strain. Appl. Environ. Microbiol. 58: 1089 1094. Wong HH (1996) Modelling studies of the interaction between homogenization, centrifugation and inclusion bodies. PhD dissertation, University of Adelaide. Wong HH, ONeill BK and Middelberg APJ (1996) Centrifugal recovery and dissolution of recombinant Gly-IGF-II inclusionbodies: The impact of feedrate and re-centrifugation on protein yield. Bioseparation 6: 185192. Wong HH, ONeill BK and Middelberg APJ (1997) Cumulative sedimentation analysis of Escherichia coli debris size. Biotechnol. Bioeng. 55(3): 556564.

Address for correspondence: A.P.J. Middelberg, Department of Chemical Engineering, University of Cambridge, Pembroke Street, Cambridge CB2 3RA, U.K.