Albumin denaturation during ultrafiltration - Wiley Online Library

Albumin Denaturation During Ultrafiltration: Effects of Operating Conditions and Consequences on Membrane Fouling Martine Meireles, Pierre Aimar,* and Victor Sanchez (URA CNRS de Ghnie Chimique) L aboratoire de Ghnie Chimique et Electrochimie, Universith P. Sabatiec 37062 Toulouse Cedex, France Received January 16, 1990Accepted January 29, 7991

Ultrafiltration of high-purity grade bovine serum albumin has been carried out under various temperatures between 5 and 30°C and at various cross-flow velocities, pressures, and concentrations with the aim of studying protein denaturation and its consequences on the process. Three different pump heads have been tested. Denaturation of proteins in solution has been monitored by laser light scattering and size exclusion chromatography. The rate of protein denaturation increases with temperature, cross-flow,and time. It is observed that membrane fouling is different whether denaturation has occurred or not. Under high-concentration polarization, denaturation can occur in the boundary layer if the wall concentration exceeds 400 g/L. It is shown how the residence time, operating temperature, and pressure play an important part in membrane fouling. This can provide guidelines for process design and control. Key words: albumin denaturation ultrafiltration protein denaturation membrane fouling INTRODUCTION

Product quality in downstream processing is of increasing importance with respect to the high value and specific bioactivity of molecules. Procedures used to control denaturation are often based on empirical knowledge. Damage to proteins such as alteration of the structure, losses in bioactivity, or even breakage is often observed at both lab~ratory”””’~ and industrial scale ultrafiltration, although its origins are not well known. Recent results with globular proteins show that damage due to shear stress is much less severe than expected whether in Couette f l o ~ , ~ ~stirred , ’ ~ , ’ tank, ~ or pumps.15 Considering the possibility that other effects are likely to be responsible for the apparent sensitivity to mechanical shear, some authors have searched other origins of protein damage. For example, the influence of high velocity gradients in the vicinity of solid-liquid and gas-liquid interfaces (e.g., gas bubbles appearing during flow recirculation) was studied by Narendranathan and Dunnillg and Fink and Rodwell.’ Stephen and Gangulin” showed that heating a P-lactoglobulin solution above 50°C induces an increase in solution turbidity reflecting protein aggregation. Harwalkar: who high* To whom all correspondence should be addressed

Biotechnology and Bioengineering, Vol. 38,Pp. 528-534 (1991) 0 1991 John Wiley & Sons, Inc.

lighted the role of temperature on the structural damage, observed that when the protein concentration is increased, denatured proteins tend to aggregate and form either a precipitate or a gel. The same observations were reported by Friedli et a1.’ using bovine serum albumin solutions. In this case, heating the solution above 50°C leads the proportion of polymers to increase and the solution to gel. The influence of protein concentration has also been investigated. In the case of bovine serum albumin, Lambin et al.’ observed that a 30% protein solution gels at 30°C. The scope of .this article is to investigate the importance of those thermal and mechanical protein denaturations on ultrafiltration performances. The influence of temperature, recirculation flow rate, transmembrane pressure, or protein concentration is examined using bovine serum albumin as test protein. Possible consequences of the conclusions on design, choice of operating conditions, and process control are discussed in the last section. The purpose is to stress the importance of protein denaturation in general. Therefore, albumin has been chosen as a model protein. Although we can reasonably expect that the conclusions can be generalized to other proteins, it must be kept in mind that any biomolecule has a specific sensitivity to the various operating parameters. EXPERIMENTALS Test Proteins

High-purity (99.8%) grade bovine serum albumin solutions (120 g/L, pH 6,8) were purchased from Institut Merieux (Marcy l’Etoile, France). Test solutions were prepared by dilution of the above in distilled water and addition of NaN3(1 mg/L) to prevent bacterial contamination. The concentrations of the test solutions were measured with a UV photometer ( A = 278 nm). The turbidity of the retentate was monitored with a Hatch Ratio 2000 turbidimeter so as to detect the presence of particles. Size exclusion high-performance liquid chromatography (HPLC) (TSK G3000SW column)

CCC 0006-3592/91/050528-07$0400

and laser light-scattering analysis (Coulter N4D) were also used for the detection of albumin polymers and of protein aggregates. Figure 1 shows typical results obtained with the initial high-purity solution by these characterization techniques. Ultrafiltration Device and Procedure

A UFP-2 module and polysulfone IRIS 3026 membranes were supplied by Tech-Sep (Miribel, France). Nominal molar weight cutoff (MWCO) membranes of 10,40, and 100 kD were used. The rejection coefficients of bovine serum albumin after fouling were typically 100,100, and 98%, respectively. The following procedure was applied to any new membrane: (i) cleaning of the membrane to remove traces of solvent left over from the manufacturing process and (ii) ultrafiltration of distilled water under 300 kPa until the permeate flux became constant. Temperature was controlled within 0.5"C. The permeate flux was measured with a Mettler electronic balance logged to a personal computer (Epson). The permeate was periodically recycled in the feed tank so as to maintain a constant bulk concentration. All the experiments were performed with 2 L of solution. Fouling was char-

acterized by the difference in permeability (or hydraulic resistance) measured before and after a run. Although shear stresses can be important at various points of an ultrafiltration device, the pump head is a piece of equipment where high shear forces can be expected. Protein damage was therefore studied with three different types of pumps: Gear pump: PM 10-20, PCM Moineau, (Vanves, France). Centrigugal pump: U62B1, Micropump Corp. (Concord, CA). Screw pump: P2 MG1, PCM Moineau, (Vanves, France). RESULTS Influence of Temperature

The conditions were 100 kPa, 1 m/s, and 8 g/kg. Figures 2a, 2b show the variations in retentate turbidity 35 30

25

t! 1.5

5

.-x .E e z

1 .o

20

15 10

5 0.5

0 0

20

60

40

0.0

80

Time (min) (a)

Elution time ( min ) (a)

r

0

20

40

60

80

100

120

140

Time (min) 1

10

100

1000

Size (nm) (b) Figure 1. Size exclusion HPLC (a) and laser light scattering (b) spectra of high grade bovine serum albumin. I monomers, 99.8%.

(b) Figure 2. (a) Variations in turbidity with operating temperature: (m) 5°C; (A) 8°C; (0)15°C; (A) 22°C; (0)30°C. (b) Variations in filtration flux with operating temperature: (0) 8°C; (A) 22°C. The transmembrane pressure is set at 100 kPa and the feed recirculation rate and concentration are 1 m/s and 8 g/kg, respectively.

MEIRELES, AIMAR, AND SANCHEZ: PROTEIN DAMAGE DURING ULTRAFILTRATION

529

and in filtration flux at 5,8, 15, and 22°C. Below 8°C no significant variation in turbidity is observed while the flux reaches a steady level. Conversely, the turbidity continuously increases during the run for temperatures above 8°C. A straight line would suggest that the denaturation occurs at a constant rate which is not the case at 22 or 3WC, as seen in Figure 2a. At 22°C the flux does not reach a stationary value, and after 80 min, it is lower than the one obtained at the lowest temperature. The analysis by size exclusion HPLC and light scattering of the solutions performed at this temperature show that the increase in turbidity reflects protein denaturation (Fig. 3). Polymers and rather large species, probably protein aggregates, are now present in the solution. The values of the hydraulic resistance measured after experiments at 8 and at 22°C are reported in Table I. The increase in hydraulic resistance due to fouling is more severe when denatured proteins are present in the solution.

that the flux decline observed after the first 30 min is due to the denatured fraction of the proteins. Influence of the Transmembrane Pressure

The pressure was varied from 200 to 350 kPa at 8°C using a gear pump. For each pressure the time to reach a steady flux depends on the recirculation rate and on

I

Influence of the Pump Head 20

15

Figure 4 shows the variations in turbidity measured at 8 and at 22°C using different pump heads. A significant increase in turbidity is observed at 22°C whatever the system, with only a slight advantage to be gained with the gear pump. At the lowest temperature, no variation in turbidity is observed during the 80-min experiments whatever pump was used.

25

Retention time ( min) (a)

Influence of the Recirculation Rate

The same gear pump as above was used to generate three cross-flow velocities: 0.5, 1, and 1.5 m/s. The turbidity and filtration flux measured at 8°C and 22°C are plotted in Figure 5. At 8"C, the solution turbidity does not change, whatever the recirculation rate while a significant increase is observed at 22°C. The analysis performed on the retentate shows a protein damage quite similar to those of Figure 3. Denaturation is faster at high recirculation rates. Figure 6 shows the variations in flux under the same conditions. The hydraulic resistances measured after 80 min are reported in Table I. Again these results support the view that fouling is more severe when a solution contains denatured proteins and

Particle size (nm) (b) Figure 3. Size exclusion HPLC (a) and laser light scattering (b) spectra of turbide bovine serum albumin solutions. I monomers, 94.4%; I1 polymers, 5.3%.

Table I. Hydraulic resistance of an IRIS 100-kD membrane before and after runs at various cross-f lows and temperatures.

Membrane Clean Fouled after experimental run at 8°C Fouled after experimental run at 8°C

530

Cross-f low (m/d

Hydraulic resistance (10" m-')

0.5 1.o 1.5 0.5 1.o 1.5

4.65 15.6 15.6 15.6 17.5 28.5 34.4

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 38, NO. 5, AUGUST 1991

30

c

l5

t

0 1 0

I

I

I

I

20

40

60

80

i

0

20

40

60

80

Time (min)

Time (min)

h

5 c:

l o

s

0

20

40

60

80

Time (min)

0

I

I

I

I

Figure 4. Variations in turbidity at 8°C (open symbols) and at 22°C (closed symbols) with three types of pumps: (A) screw pump, (w) centrifugal pump; (+) gear pump.

the feed concentration. No variation in turbidity is observed during the set of experiments. The variations in steady flux are shown in Figure 7. These curves have the typical shape (limiting flux) usually reported,6.*”’ except in two cases for which the filtration flux decreases at high pressure. This is observed at a low recirculation rate for a given concentration and at a high concentration for a given recirculation rate. Such a trend has already been reported by Jonsson’ with dextran and whey proteins at 90 g/kg and high pressure (2000 kPa). In addition, other experiments have been performed using 10- and 40-kD membranes. The hydraulic resistances of the clean membranes were 4.3 x 10” and 1.3 x 10l1m-’, respectively. The temperature was 8°C. For the tighest membrane, the curves have a regular shape. For the medium one, a flux decline is observed above 28 kPa (0.5 m/s; 40 g/kg). When the limiting flux decreased at high pressure, a deposit was visible on the membrane after the run. The deposit was removed from the surface by vigorous rinsing with distilled water and then dissolved in an acetate buffer. This solution was analyzed by size exclusion chromatography (TSK G3000 SW column) showing that polymers of albumin represented 25% w/w of the total protein content of the solution.

DISCUSSION Whereas previous works have shown that adsorption of proteins on the membrane play an important part in the earlier stages of fouling, the experiments presented in this article offer the conclusion that long-term fouling in ultrafiltration of albumin is tightly related to the denaturation of the proteins. Although it is quite certain that adsorption cannot be avoided, it can be limited by a proper choice of the membrane material and solution pH and ionic strength. On the contrary, the parameters to be adjusted to eliminate long-term fouling are different and belong to the process engineering part of the problem more than to the physical biochemistry one. As considered here, denaturation i.e., aggregation and polymerization, is the result of a combination of a variety of parameters, among which temperature, crossflow, concentration, and time have been identified as significant. There are most probably as many levels of denaturation as combinations of the above parameters, which might explain some of the inconsistencies reported in the literature and the relative slow pace at which the fouling science is progressing.


531

0

40

20

60

80

100

120

140

Time (min)

ure 2a. For instance, if an increase of 25% in turbidity is the upper limit acceptable in terms of product quality or ultrafiltration performances, the hatched region in Figure 8 corresponds to conditions of time and temperature which would lead the turbidity to exceed the threshold. Denaturation would then be avoided by a convenient choice of operating conditions such that the point defined by the temperature and the residence time lays out of the hatched region. We now examine what would be the consequences of a change in the operating temperature. In a continuous ultrafiltration plant, the average residence time 7 depends on the total holdup volume V and on the output flow Qo: 7

(1)

= V/Qo

Assuming a 100% rejection coefficient and defining the concentration factor X as the ratio of the final concentration over the initial one allows us to get Qo from the mass balance equation Qo = AJ/(X - 1) 0

40

20

60

80

100

120

140

whereA is the membrane area and J the flux. Substituting for Q0 in Equation (1) would give T as

Time (min) r=--

Figure 6. Variations in filtration flux with recirculation rate at 8°C (open symbols) and at 22°C (closed symbols): (A) 0.5 m/s; (0)1.0 m/s; (0)2.0 m/s (solution concentration 8 g/kg).

or 0

I

1

2

vx-1

A

I

I

3

4

(2)

(3)

J

As an increase in temperature generally generates a flux increase, r is expected to become smaller when the processed fluid is heated. As depicted in Figure 8, the sensitivity of the flux to the temperature variations can lead to different situations. If the residence time reduction produced by the flux increase is not sufficient (line a, Fig. 8), the increase in temperature ends up with inacceptable protein denaturation. This could be avoided by a better design, e.g., by improving the ratio V/A so as to keep the operating point off the hatched region (line b, Fig. 8). This illustrates how changing the temperature may require a modification in the process design. However, we must note that since large aggregates are likely to be rejected by ultrafiltration mem-

Transmembrane pressure AP ( lo5 Pa)

Figure 7. Variations in filtration flux with transmembrane pressure for various recirculation rates and solution concentrations: (H)1 m/s, 40 g/kg; (*) 1 m/s, 20 g/kg; (0)1 m/s, 8 g/kg; (A)1.5 m/s, 20 g/kg; (+) 2 m/s, 20 g/kg. Temperature is set at 8°C.

These remarks underline the fact that modern membrane process engineering should give the same importance to the pump, the volume and design of pipes and of valves, and the temperature as to the choice of the membrane. Two examples are given in the following discussion to illustrate the consequences of different strategies in design and process control. For each flow rate and pumping system, contours of constant denaturation (i.e., turbidity) can be plotted in time-temperature coordinates using data from Fig-

532

O

L

0

~

~

10

"

20

*

~

~

30

'

.

40

Temperature ("C)

Figure 8. Data of Figure 2a plotted in time-temperature coordinates. The hatched region corresponds to an increase of more than 25% in turbidity.


~

branes, they tend to concentrate in concentration loops, thus adding another time parameter (the run duration) to the residence time. The experiments reported here underline the importance of other damaging effects which result, for high pressure, in a decrease in filtration flux accompanied by a denaturation behavior. The apparent effects of high pressures could be brought together with high concentration effects, especially since this behavior is observed for experimental conditions corresponding to the highest levels of polarization (low recirculation rate or high bulk concentration). Since the present gelling behavior is not observed every time a limiting flux is obtained with a nondenatured solution, we focused our interest on the concentration polarization associated with these observations. In order to estimate the value of the wall concentrahas been used as a link betion, the osmotic tween the applied pressure and the wall concentration. The following relationship was used for the flux: J = (AP

- An)/j.d?h

(4)

where R h is the hydraulic resistance of the fouled membrane and An, the osmotic pressure difference, is given by

An = u ~ C +, a2Crn2+ u3Crn3

(5)

a 2 = 2.4 x and u 3 = with u l = 3.6 x 4.8 x The hydraulic resistances R h , measured after the run, are 1.6 x lo", 4.0 x lo", and 5.6 x loi1m-' for the loo-, 40-, and 10-kD membranes, respectively when no deposit is observed on the surface. For each experiment, the wall concentration is estimated from the experimental values of AP and J together with Equations (4) and (5), assuming that the solution viscosity is equal to that of the solvent. The values of the pressure and of the wall concentration corresponding to a limiting fluxJli, are reported in Table 11. The wall concentration corresponding to Jlimis always around 300 g/kg for any set of conditions and any membrane. Values of the wall concentration estimated from runs where a deposit was found are also reported in Table I1 for the 100- and 40-kD membranes. This concentration is about 400 g/kg. No analysis could Table 11. Wall concentrations estimated from Equations ( 5 ) and (6) for transmembrane pressure corresponding to (a) limiting flux and (b) visible deposit.

be made on high-purity solutions of 400 g/kg to help to understand the meaning of this limit of 400 g/kg. For the 10-kD membrane, a limiting flux decline was never observed, and accordingly the calculated concentration never reaches 400 g/kg within the range of operating conditions investigated. These results suggest that the denaturation observed at high pressure is associated with an increase in wall concentration above a yield value (here 400 g/kg). They can be put in relation with the high proportion of polymers (25%) reflected by HPLC analysis. Studying bovine serum albumin, Friedly et a1.2 observed a relationship between the increase in proportion of polymers induced by heating and the deposition of denatured proteins. In the present work, such an explanation could be convenient assuming that the increase in wall concentration leads to the polymerization or gelation of the solution above 400 g/kg. These findings could have consequences on the performances of plants used to concentrate protein solutions. According to the film model, the wall concentration depends on that of the bulk, and it is, therefore, expected to increase during a concentration run if the operating conditions are kept constant. Constant wall concentration lines are drawn in Figure 9. A limiting flux is expected beyond the 300 g/kg line, whereas the hatched region corresponds to wall concentrations higher than 400g/kg and should then be avoided in order to prevent denaturation associated with gelling behavior and flux decline. equation C,,, = 300 glkg

equation C,,, = 400 gikg

\

/

5.0

.3?

$ 2L

5til

4.0

3.0

C* w-

I

a-

2.0

0 0

g;

a-

1 .o

0 0 0.0

0.5

1.0

1.5

2.0

2l

Transmembrane pressure AP

3'0

Operating pressure

p1

100

100 100 100 40 10

U

co

(m/s) 1 1 1 0.5 0.5 0.5

(g/kg)

ASirn, 10' Pa

C,(a) (dkd

Cm(b) (g/kg)

8 20 40 40 40 40

1.9 1 .5 1.3 1.1 1.2 1.8

305 300 297 308 298 301

407 402 396 411 398 -

Figure 9. Example of the variations in flux versus transmembrane pressure for two feed concentrations Ciand Cp The hatched region corresponds to the conditions for which a gelling behavior and flux loss can be observed (wall concentration more than 400 g/kg). Two precedures can be used to concentrate from Ci to C f : if pressure is maintained constant at P1, the wall concentration is expected to increase beyond 400 g/kg, line a. Line b represents a pressure monitored so as to keep the wall concentration at a constant value (here 300 g/kg).


533

If the concentration from Cito C, is performed at constant pressure, the line a in Figure 9 shows that the wall concentration gradually increases and is expected to exceed 400 g/kg sooner or later and hence to reach conditions of denaturation, whereas the initial conditions were safe. Another option represented by line b shows how denaturation has a better chance to be avoided with a pressure properly monitored so as to keep the wall concentration at a constant value, (of course, below 400 g/kg), the flux for each level of concentration being the same as in the previous case (line a). The second option allows the deposit to be avoided. From a practical point of view, the equation of line b is obtained by a combination of Equations (4) and (5):

AP = AII(300) + JpRh

(6)

The hydraulic resistance Rh of the fouled membrane should be measured after runs on a pilot plant.

temperature, concentration, time) should be considered as the tight link between the membrane module itself and peripheral equipments such as pumps, tanks, heat exchangers, pipes, etc. Financial support by Tech-Sep (01703 Miribel, France) is gratefully acknowledged, as well as engineering assistance by J. P. Lafaille.

NOMENCLATURE membrane area (m2) bulk concentration (g/kg) membrane (or wall) concentration permeate flux (kg/mz s) or (m’/m’ s) pressure-independent limiting flux permeate viscosity (Pa . s) transmembrane pressure (Pa or bar) osmotic pressure difference fouled membrane hydraulic resistance (m-’)

References

CONCLUSION

In ultrafiltration, once the adsorption process has reached an equilibrium, proteins such as albumin are not very foulant in nature but can become so if the operating conditions allow their partial denaturation. Further, in the absence of denaturation checked by size exclusion chromatography, no flux decline was observed after the first few minutes. Denaturation seems to be triggered by the temperature: Below a threshold (8°C here) no aggregation was observed in any condition. Beyond it, the amount of denatured protein increased with time, temperature, and cross-flow. Although the gear pumps seem less damaging, the type of pump head does not appear here as crucial as the temperature or cross-flow. The apparent effects of pressure on denaturation seem to be explained by an excess of concentration in the boundary layer. From the viewpoint of the process design, the sensitivity of a protein to the operating parameters (pressure,

534

Rodwell, , V. W. 1975. Biotechnol. Bioeng. 17 1029. 1. Fink, D. .I. 2. Friedli, H., Fournier, E., Volle, T., Kistler, P. 1976. Vox Sang. 31: 283. 3. Goldsmith, R. L. 1971. Ind. Eng. Chem. Fund. 10: 113. 4. Harwalkar, V. R. 1986. Milchwissenschaft 41: 4. 5. Jonsson, G. 1984. Desalination 51: 61. 6. Lafaille, J. P., Sanchez, V., Mahenc, J. 1984. Entropie 118: 29. 7. Lambin, P., Rochu, D., Herance, N., Fine, J.M. 1982. Rev. Franc. Transfus. Sang. 23: 5. 8. Michaels, A. S. 1968. Chem. Eng. Prog. 64: 31. 9. Narendranathan, T. J., Dunnill, P. 1982. Biotechnol. Bioeng. 24: 2103. 10. Probstein, R. F., Shen, J. S., Leung, W. F. 1978. Desalination 24: 1. 11. Rhee, B.H., McIntire, L.V. 1986. Chem. Eng. Commun. 47: 147. 12. Shephen, J., Gangulin, G. 1979. Milchwissenschaft 34: 3. 13. Thomas, C. R., Nienow, A. W., Dunnill, P. 1979. Biotechnol. Bioeng. 21: 2263. 14. Twineham, M., Hoare, M., Bell, D. J. 1984. Chem. Eng. Sci. 39: 509. 15. Virkar, P. D., Narendranathan, T. J., Hoare, M., Dunnill, P. 1981. Biotechnol. Bioeng. 23: 425.
