Disc Stack Centrifugation Studies

MCI strain cell harvesting, cell washing, debris removal and precipitate separation by intermittent discharge disc stack centrifugation were examined. In all cases, a Westfalia SAOOH 205 intermittent discharge disc stack centrifuge (Westfalia Separator Ltd., Milton Keynes, U.K.) fed via a type SB14R Monopump (Monopumps Ltd., London, U.K.) was used. Unless otherwise stated, the centrifuge was operated with a full stack o f 43 discs and a rotational speed o f 1 0 , 0 0 0 rpm giving a sigma factor (Z) equal to 1560 m^ (refer to Section 7.2.2 for a description o f the sigma factor). The full scale centrifuge bowl volume and solids holding space were 0.6 L and 0.3 L respectively.

During the experimental trials only full discharges were used with the optimum discharge time for each material determined from optical density measurements at 670 nm (Section 3.2.3). When the ratio o f supernatant to average feed optical density equals 1 overflow o f the particulate phase into the supernatant stream is indicated due to the filling

o f the centrifuge solids holding space. The optimal discharge time was therefore taken when this ratio equalled 0.8. The feed pump was stopped before each discharge to limit the loss o f feed material from the bowl.

In order to apply centrifugation models the suspension density (Section 3.2.8.1), the particle density (Sections 3.2.8.2 and 3.2.8.3), the suspension viscosity (Section 3.2.8.4), aqueous fraction (Section 3.2.3.1) and dry weight (Section 3.2.2.2) of each feed material were determined.

3.4.3.1 Cell Harvesting and Washing

For cell harvesting studies, MCI strain fermentation broth from 30 L fermentations and a 1000 L fermentation was used (Section 3.3.2.3) and processed 24 h after inoculation at flowrates o f 100, 200 and 400 L.h '. Washing o f the cell concentrate was also carried out by diluting the recovered cells 1:5 with 100 mM potassium dihydrogen phosphate buffer at pH 6.5 and recentrifuging at flowrates o f 20 L.h'' and 60 L.h '.

For both stages and at each flowrate, supernatant stream samples were taken at regular intervals during centrifuge operation and were analysed for optical density at 670 nm (Section 3.2.3) and microtube dry weights (Section 3.2.2.2). Steady state samples, i.e. samples collected after 10 bowl volumes had passed through the centrifuge (Clarkson et a i, 1993), were analysed for particle size using the Elzone method (Section 3.2.7.1) with recovery efficiencies and grade efficiencies being determined (Section 3.4.3.4). Solids concentrations were also measured for the final discharged solids and supernatant using the microtube dry weight method (Section 3.2.2).

3 4.3.2 Debris Removal

To examine debris removal, disruption o f the washed cell slurry was carried out using a Manton Gaulin LAB-60 high pressure homogeniser (APV Baker, Crawley, U.K.) for material prepared in the 42 L fermenter, and a Manton Gaulin K3 high pressure homogeniser (APV Baker, Crawley, U.K.) for material prepared in the 450 L and 1000 L fermenters (refer to Section 3.4.3.1 for the harvest and wash procedure). In both cases 5 discreet passes and an operating pressure o f 500 barg was used. To avoid protein

dénaturation and enzyme inactivation, the temperature o f homogenate was maintained at 4°C by glycol cooling.

Grade efficiency curves were determined for the centrifugation o f homogenised cells from 30 L fermentations. Due to the small volume o f material available, experiments were carried out by both diluting the homogenate to 1% (w/v) and by scaling down the centrifuge to only 9 active discs (Maybury, 1997). The diluted homogenate was processed at flowrates o f 200, 400 and 600 L.h'% while the scaled down centrifuge was operated at 5 and 10 L.h'' so as to maintain the same Q /I ratio as the full scale unit operating at flowrates o f 25 and 50 L.h ' respectively.

In the case o f the scale-down unit, 3 blanking inserts were used and also 1 bottom insert to raise the active discs above the turbulent region at the bottom o f the centrifuge bowl (Maybury, 1997). This resulted in an approximately 45% reduction in the bowl volume and solids holding space. Steady state samples, i.e. samples collected after approximately 10 bowl volumes had passed through the centrifuge (Clarkson et a l,

1993), were analysed for particle size using the Elzone method (Section 3.2.7.1) and recovery and grade efficiencies determined (Section 3.4.3.4).

Steady state recovery efficiencies for a full scale disc stack at the process concentration were determined at flowrates o f 25, 50 and 100 L.h ' using homogenised cells from 450 and 1000 L fermentations and compared to predicted recovery efficiencies based on the smaller scale results.

The effect o f high solids loadings on overall recovery efficiency due to solids overflow was examined using homogenised cells from 30 L fermentations processed with the scaled-down disc stack. In both cases, the recovery efficiency, E^, was determined from Elzone particle size data (Section 3.4.3.4). Solids concentrations o f the pooled discharged solids and supernatant were also measured using the microtube dry weight method (Section 3.2.2.2). ADH and protein assays (Sections 3.2.1.1 and 3.2.1.2 respectively) were carried out on the feed material and supernatant to determine the yield o f soluble ADH and protein.

3.4.3.3 Precipitate Separation

The separation o f precipitate after each cut point was examined as part o f ftill process runs (Section 3.4.4). A feed flowrate o f 50 L.h ' was used in the large pilot scale run, while in the small pilot scale run a scaled-down disc stack was used at flowrates o f 5 and 10 L.h'' in order to give equivalent processing to a full stack at 25 and 50 L.h ' respectively.

Because the precipitate particle size was below 1 pm and laser sizing was used (Section 3.2.7.2), the precipitate PSD was only described in relative terms and grade efficiency curves could not be determined. The efficiency o f precipitate separation was therefore determined from feed material and supernatant optical densities at 670 nm (Section 3.2.3) and also microtube dry weights (Section 3.2.2.2) with correction to account for the mass o f any ammonium sulphate present.

3.4.3 4 Recovery and Grade Efficiency Determination

Grade efficiencies (refer to Section 7.2.3 for definition and description) for the processing o f a particular material was determined from Elzone particle size data (Section 3.2.7.1) using the approach o f Mannweiler (1989).

Results from the Elzone particle sizer were expressed on a cumulative volume basis as a function o f particle size logarithmically divided into 128 size channels. To reduce error and give a degree o f smoothing to the data, the 128 channels were further reduced to 26 wider size channels. Assuming particles in the feed and supernatant have the same constant density, the recovery efficiency (alternatively known as the mass yield), E^, was firstly calculated as:

E ( = 1 - ^

(equation 3.8) with both Vf, the total volume o f solids in the feed, and v^, the total volume o f solids in the supernatant, being determined from the respective volume distribution. Grade efficiencies, T(d), were then obtained from the equation:

(equation 3.9) Both AFs(d), the centrifuge supernatant particle size distribution, and AFf(d), the centrifuge feed particle size distribution, are expressed as 'percent in range' distributions (i.e. the sum over all size fractions is 1 00 %) and were obtained by calculating the difference between two values in the respective volume distribution for consecutive particle sizes d, and d2=d,+Ad, i.e:

AF(di,d2) = F(dj) - F(dz;

(equation 3.10) Grade efficiencies were obtained in the form o f discrete grade efficiency values based on the arithmetic mean particle size (d,+d2 ) / 2 for each size interval, which were plotted as a function o f the dimensionless particle size (d /d j, where d^. is the critical diameter. Numerical analysis o f experimental grade efficiency data was carried out using the ORIGIN software package (Section 3.5).