Biotechnol. hog. 1992,8, 391-396
391
Screening of Insect Cell Lines for the Production of Recombinant Proteins and Infectious Virus in the Baculovirus Expression System T. J. Wickham,tJ T. Davis,$ R. R. Granados3 M. L. Shuler,**tand H. A. Woods School of Chemical Engineering, Cornell University, Ithaca, New York 14853, and Boyce-Thompson Institute for Plant Research, Tower Road, Ithaca, New York 14853
Eight cell lines derived from the insects Spodoptera frugiperda, Trichoplusia ni, Mamestra brassicae, and Estigmene acrea were evaluated for recombinant P-galactosidase and infectious virus production following infection with the baculovirus Autographa californica multiple nuclear polyhedrosis virus (AcMNPV). Production was assessed on a specific (per cell and per microgram of uninfected cellular protein) and on a volumetric (per milliliter) basis. Cell density was found t o be an important factor in comparing the cell lines due to a density-dependent inhibition of specific protein and virus production that appeared to result from cell-cell contact. After infection of cells at low-density specific fl-galactosidase production per cell would drop between 3- and 6-fold in five of the eight cell lines when plated on tissue culture plates at near-confluent and confluent cell densities. The cell lines Sf 21 and Sf 9 were least sensitive t o cell density. After accounting for cell density effects and differences in cell size, two cell lines, BTI T n 5B1-4 and BTI TnM, were identified that were superior to the other cell lines, including Sf 21 and Sf 9, in ,&galactosidase production. Optimal volumetric and specific P-galactosidase production from T n 5B1-4 and T n M cells was 2-fold and &fold higher, respectively, in both cell lines than the optimal production from Sf 9 or Sf 21 cells. The T n 5B1-4 cell line also had the highest viability of all the cell lines at 3 days postinfection and could be adapted to serum-free media. Specific (per microgram of cellular protein) infectious virus production between cell lines varied by less than 2-fold between cell lines after cell size differences were taken into account. No correlation was found between recombinant protein and NOV production although both were found to decrease similarly with increasing cell density.
Introduction The baculovirus expression vector (BEV) system has been widely used for the production of an array of proteins for research and industry (Luckow, 1990). The system is capable of producing very high levels of recombinant protein, approaching 50% of the total cellular protein following infection of insect cells with recombinant virus in some cases. For most of the proteins studied, the BEV system produces biologically active proteins possessing many of the posttranslational modifications present in the native form of the protein. Insect cells are capable of performing 0-linked and N-linked glycosylation, phosphorylation, myristylation, palmitation, and cleavage of certain proteins to their active forms. For example, the BEV system has been shown to produce and secrete propapain, which requires glycosylation and proteolytic cleavage of the prepro form of the protein for activity (Vernet et al., 1990). Although differences exist in Nlinked glycosylation between insect and mammalian cells, the full extent and consequence of these differences have not yet been fully determined. However, most of the proteins produced thus far retain full biological activity and/or antigenicity (Luckow, 1990). Very high levels of some recombinantproteins have been achieved that approach the levels of polyhedrin (1000 mg/ L); however, the BEV expression system, in most cases, + Cornell University.
*
Present address: The Scripps Research Institute, 10666N. Torrey Pines Road, La Jolla, CA 92037. Boyce-Thompson Institute for Plant Research. 8756-7938/92/3008-0391$03.0010
has produced at least 2 orders of magnitudes lower levels of glycosylated, secreted proteins in Spodoptera frugiperda-9 (Sf 9) cells (Luckow, 1990). Erythropoietin concentrations achieved using the BEV system have been reported at 10 mg/L (Wojchowski et al., 1987) while levels of tissue plasminogen activator (Steiner et al., 1988), an extracellular domain of complement receptor-2 (Moore eta l., 19911, and interleukin-2 (Smith et al., 1985) have only reached concentrations of 1-3 mg/L. These lower production levels of secreted proteins have limited the use of the BEV system in industry due to the use of mammalian cell expression systems that can effectively match or better the above concentrations in many instances. Also, these lower levels impede structure/function mutational studies of receptors in research where large quantities of the mutant protein must be produced in order to obtain crystals for X-ray structure determinations (Greenfield, 1988). However, it is important to note that nearly all of the proteins reportedly produced using the BEV system have used only the Sf 9 cell line. One explanation for the discrepancies between the expression levels of certain proteins could be differences in the construction of the expression vector. However, another possibility is that Sf 9 cells, or possibly, insect cells in general, are limited a t some point in their modification and/or secretion capacity. Sf 9 cells have been found to possess rate-limiting steps in the posttranslational processing and secretion of certain proteins (Jarvis and Summers, 1989; Vernet et al., 1990). Thus, the identification of cell lines capable of higher rates of translational and posttranslational processing would be
0 1992 American Chemical Society and American Institute of Chemical Engineers
Biotechnol. Prog., 1992, Vol. 8, No. 5
392
highly desirable in the production of many proteins using the BEV system. Work with mammalian expression systems has shown that the choice of a cell line can be critical to obtaining active quantities of secreted proteins (Whang et al., 1987). A large number of insect cell lines have been established, and the procedures for isolating new cell lines have been simplified (Granados et al., 1986). Hink et al. (1991) have previously established the importance of evaluating cell lines for the production of secreted and nonsecreted proteins for a number of cell lines. They found that individual cell lines differed in their ability to synthesize secreted versus nonsecreted proteins, presumably due to the differing machinery that each cell line possessed. A few cell lines were identified that produced up to 3 times more human plasminogen than Sf 9 cells; however, none were superior to Sf 9 in @-galactosidaseproduction. King et al. (1991),however, reported a Mamestra brassicae cell line that could synthesize twice as much @-galactosidase per milliliter as Sf 9 cells when both were infected under optimal conditions. In the present paper, a screening procedure is reported that is used to evaluate eight insect cell lines, including four novel cell lines, for their ability to produce @-galactosidaseand infectious virus. The effect of cell density was included in the comparison on the basis of the previous finding that polyhedrin and nonoccluded virus production from attached Sf 21 and Tn 368 cell lines plated at high cell density on tissue culture plates was decreaseddue tocellcontact inhibition (Wood etal., 1982). This sensitivity indicated that cell density was also a critical factor in reliably evaluating the inherent recombinant production capabilities of different cell lines using the BEV system.
Materials and Methods Cell Lines, Virus, and Media. Trichoplusia ni BTIT n M is a novel cell line (Granados et al., manuscript in preparation) established from T. ni midgut tissue. Two other novel cell lines, BTI-Tn 5Bl-4 (Granados, unpublished) and BTI-Tn Ap2 (Granados et al., manuscript in preparation), were established from T.ni eggs. BTI-Ea 88 (Granados et al., manuscript in preparation) is an attached cell line derived from Estigmene acrea BTI-EaA (Granados and Naughton, 1976). Other cell lines used were T n 368 (Hink, 19701, IPLB-Sf 21AE (Vaughn, 19771, Sf 9 (Summers and Smith, 1987), and M. brassicae Mb 0507 (King et al., 1991). These cell lines were grown in TNM-FH medium (Graces insect cell medium (Gibco, Grand Island, NY) plus lactalbumin and yeastolate) supplemented with 10% FBS (Hyclone, Logan, UT) (complete medium). The recombinant Autographa californica multiple nuclear polyhedrosis virus (AcMNPV) containing the Escherichia coli @-galactosidasegene fused to the polyhedrin gene (E2-@-gal)was obtained from MicroGenesys (Stamford, CT). The nonrecombinant 1A strain (Wood, 1980) was used to measure nonoccluded virus production. Analysis of the Size of Cell Lines. The relative sizes of the cell lines were assessed by two independent methods. The size distribution of each cell line was measured on a Coulter counter (Coulter Electronics). From the size distribution, the mean cell size was obtained. The relative sizes of the cell lines were also assessed by measuring the amount of protein per cell using a protein assay kit (BioRad). A cell sample was first counted in a Coulter counter to determine the cell concentration and then spun at lOOOg for 5 min to pellet the cells. The pellet was washed twice with phosphate-buffered saline and then resuspended in
a known volume of PBS. The sample was then sonicated to disrupt the cells. The protein concentration in each sample was then measured using a standard Bradford total protein assay kit (Bio-Rad), and this value was then used to determine the amount of protein per lo6 cells. Assay for Nonoccluded Virus Production. Nonoccluded virus production was measured by seeding 24 multiwells with 5 X lo4 viable cells/well of each cell line and then adding nonrecombinant, 1A strain of AcMNPV at a multiplicity of infection of 10 plaque-forming units (pfu)/ cell in a final volume of 0.7 mL. The multiwell plates were then centrifuged at l00Og for 1h (0 h postinfection (pi)). The medium from each well was then removed and replaced with 0.5 mL of fresh medium. At 4 days pi, the medium was removed and assayed for plaque-formingunits per milliliter using plaque assay (Wood, 1977). Recombinant and nonrecombinant virus inocula were also titered using plaque assay. Briefly, serial dilutions of samples were made and added in 1.0-mL aliquota to 6 multiwell plates containing lo6celldwell. The plates were then centrifuged at lOOOg for 1h and then overlayed with agarose (FMC, Rockland, ME). Plaques were counted under a microscope after five days. Production of @-Galactosidaseas a Function of Cell Density. Mid log phase cells in T-75 flasks were inoculated with the E2-@-gal virus at an MOI of 25 pfu/cell for 2 h. The infected cells were then immediately seeded at densities ranging from 5 X lo4to lo6 cells/2.0 cm2well in a volume of 0.7 mL. The cells were allowed to attach for 1h after which the medium was removed and replaced with 0.7 mL of fresh medium. The cells were checked for a cytopathic effect at 24 h pi to confirm the percentage of infected cells. Assay for &Galactosidase Activity. At six days pi, E2-8-gal-infected cells and medium were removed and sonicated for 5 s in eppendorf tubes to release any remaining intracellular &galactosidase. Longer sonication times had no effect on @-galactosidaseactivity. @Galactosidase activity was measured on the basis of the rate of cleavage of o-nitrophenyl @-D-galactoside(ONPG) (Sigma). Briefly, samples were added to 0.8 mL of Zbuffer and incubated at 28 "C. ONPG was added at time zero and incubated at 28 "C for times ranging between 2 and 20 min, depending on the rate of color change. The reaction was stopped with 0.5 mL of 1 M Na2C03. The stop-time was noted, and the OD420 of the solution was then measured. The following equation was used to calculate the @-galactosidaseconcentration in International Unita per milliliter using an extinction coefficient of 4.6 mL/pmol for o-nitrophenol: IU/mL = (OD,,,) (1500 pL)/ (4.5 mL/pmol)(min of incubation)(pL of sample) Production could be converted to a microgram per lo6 cells using the cell concentration in each well and a specific activity of 1.0 IU/pg for @-galactosidase.
Results &Galactosidase Production as a Function of Cell Density in Serum-Containing Medium. Eight cells lines at varying densities were infected with a recombinant E2 strain of A. californica nuclear polyhedrosis virus containing the gene for E. coli @-galactosidasefused to upstream regions of the polyhedrin gene. At 24 h pi, over 90 % of cells in all the cell lines were infected on the basis of cytopathic effect. Kinetic studies of intracellular and extracellular @-galactosidaseproduction from Sf 9 and T n 5B1-4 cells (Figure la,b) revealed that none of the @-ga-
3Q3
Bbtechnol. Rog., 1992, Vol. 8, No. 5
Table I. Optimal Production of 8-Galactoaidaaefrom Different Insect Cell Linea. p g of cell
4
2
0
protein/ 106 cells IU/W cells 306 1090 (5.0 X 104) 215 873 (5.0 X 104) 215 177 (1.0 X lo6) 204 156 (3.0 X lo6) 413 302 (1.0 X lo6) 450 717 (1.0 X lo6) 469 1155 (5.0 X 104) 236 678 (1.0 X lo6)
cell h e Tn 5B1-4 Tn M Sf 21 Sf 9 Ea 88 Tn 368 Mb Tn Ap2
6
viability IU/mL at3dpi 216 (2.5 X 106) 98 236 (4.0 X 106) 97 104 (8.0 X 106) 90 97.2 (5.0 X 106) 90 104 (5.0 X 106) 90 137 (1.5 X 105) 70 140 (1.5 X 106) 92 128 (3.0 X 106) 89
0 Celldensities(cells/2.0-cm2 well)at which maximum was reached are shown in parentheses.
Days
0
I
0
Media
I
TnSB14 Sf9
0 Sf21
A MbO507
4 2
4
6
8
Days
O.Oet0
I-
-6
2 . 5 ~ 5 5.0et5
7.5e+5
Figure 1. (a, top) &Galactosidase production kinetics in Sf 9
2 1200
cells. (b, bottom) b-Galachidase productionkineticsin Tn 5B14 cells.
2
."
1000
f
800
0 Tn 368 Tn Ap2 0 Tn M
d
lactosidase was secreted. It only appeared in the medium as a result of cell lysis which commenced a t 3 d pi. The amount of @-galactosidasepeaked at 6 d pi and remained steady with no significant degradation taking place. Cell density had a dramatic effect on the per cell production of @-galactosidasefor a t least five of the cell lines tested. All the T.ni cell lines, in addition to IZB Mb 0507, showed a t least 6-fold reductions in their specific production going from 1 X lo5 cells/2-cm2well to 8 X lo5 cells/2-cm2well. Table I shows the peak production of &galactosidase for all the cell lines on a per cell and per milliliter basis, along with the cell densities at which the peak values were obtained. A comparison between the per cell and per milliliter production values gives an approximate indication of the sensitivity of the cell line to density. Those cell lines with much lower peak production per milliliter compared to their peak production per cell were most sensitive to density. Figure 2a,b shows the effect of density on eight cell lines on a per cell basis. The production from two of highest producing cell lines, judged from their production on a per cell basis, is compared to the production of Sf 9 and Sf 21 cells in Figure 2a. The peak production on a per cell basis for the density-sensitive cell lines was at either 5 X lo4or 1 X lo5cells/well. At higher densities past the peak, production steadily declined nearly linearly with cell density up to (4-5) X lo5cells/well, where the decline then began to level off. The Tn 5B1-4 and Mb cell lines had the highest per cell &galactosidase levels of 1200 IU/106 cells and 1400 IU/106cells, respectively, which were over 7-fold higher than the per cell levels of Sf 9 and Sf 21 cells. However, the Sf 9 and Sf 21 cells were much less sensitive to cell density than the other cell lines so that their peak productions per milliliter were only about 2-fold lower than Tn 5B1-4 and Tn M cells, which were the highest in production per milliliter (Table I). The specific per cell production was independent of doubling or halving the
E
-q1 e n 5
-
1.0~6
Density (Cells/lS m m Well)
A Ea88
600 400 200
0
O.Oet0
25et5
5.0et5
7.5et.5
1.0~6
Density (Cells/lS m m Well)
Figure 2. (a, top) 8-Galactosidase production on a per million cell basis as a function of cell density for the cell lines Tn 5B1-4, Sf 9, Sf 21, and Mb 0507. (b, bottom) &Galactosidaseproduction on a per million cell basis as a function of cell density for the cell lines Tn 368, Tn Ap2, Tn M, and Ea 88.
media volume in the well while the number of cells per well constant was kept constant, which indicates that the density effect was not due to a depletion of a critical nutrient or buildup of a toxic byproduct. Production levels on a per milliliter basis are compared between the eight cell lines in Figure 3a,b. Mb cells were the most sensitive of all the cell lines to cell density so that the peak concentration of @-galactosidasereached by this cell line was reduced compared to Tn 5B1-4 and Tn M cells (Figure 3b). Because Tn 5B1-4 and Tn M cells were more sensitive to cell density, they reached peak P-galactosidase concentrations a t lower densities than the Sf cell lines, which produced much less per cell at their peak density but whose production was much less sensitive to higher cell densities (Figure 3a). The peak production levels of the Tn 5B1-4, Tn 368, Sf 9, and Sf 21 cell lines were tested on three separate occasions to check for day to day reproducibility and were found to vary by less than 10%. The amount of total protein per uninfected cell was determined to measure the relative sizes of the cells. The sizes of the cells were compared to be sure that large
1
3e4
Biotechnol. Pro& 1992, Vol. 8, No. 5
. ;
0 Tn 581-4
2
Sf9
re
3
0 Sf21
200-
A TnM
1
0
100-
n
5 0
! & , , , , , , I
O.Oe+O
]
0
O
Table 11. Production of Nonoccluded Virus from Different Insect Cell Lines
2.5e+5 5.0e+5 7.5e+5 Density (Cells/l5 m m Well)
1.0e+6
Tn368 Mb0507 0 TnAp2 A Ea88
.
3
pg
cell line
BTI-Tn 5B1-4 BTI-Tn M IPLB-Sf 21AE Sf 9 BTI-Ea 88 Tn 368 IZB Mb 0507 BTI-Tn Ap2
of protein/ 106 cella 306 215 215 204 413 450 469 236
pful cell
PfUIPi3 of cell Drotein
310 125 230 200 400 nd 320 125
1.01 0.58 1.07 0.98 0.97 nd 0.68 0.53
protein basis, the production from the cell lines was similar. Using this convention, Sf 9, Sf 21, Ea 88, and T n 5B1-4 cell lines all produced 1.0 (fO.l) pfu/pg. A t most, there were only 2-fold differences in NOV produced per microgram of cellular protein between Sf 21 and T n Ap2 cells.
Discussion
O.Oe+O
2.5e+5 5.0e+5 7.5e+5 Density (Cellsil5 m m Well)
1.0e+6
Figure 3. (a,top) &Galactosidase production on a per milliliter basis as a function of cell density for the cell lines Tn 5B1-4, Sf 9, Sf 21, and Tn M. (b, bottom) 8-Galactosidaseproduction on a per milliliter basis as a function of cell density for the cell lines Tn 368, Mb 0507, Tn Ap2, and Ea 88. differences in recombinant protein production on a per cell basis were not simply due to large differences in cell size (Table I). Sf 9, Sf 21, T n M, and T n Ap2 cells were found to contain nearly the same amount of protein per cell using a Bradford total protein assay and also to have nearly the same mean volume as assessed using a Coulter counter. Mb, Ea 88, and T n 368 cells were 2 times larger than the above four cell lines by protein and volume determinations, while the T n 5B1-4 cells were 1.5 times larger (Table I). It was determined that for every microgram of uninfected cellular protein the Tn M and T n 5B1-4 cell lines would produce 4.0 and 3.5 pg of 8galactosidase. These ratios were 5.0 and 4.5 times larger, respectively, than those of the Sf 9 and Sf 21 cells (Table I). Nonoccluded Virus Production. Nonoccluded virus (NOV) production was monitored as a function of cell density for the T n 5B1-4 and Sf 9 cell lines (data not shown). NOV production as a function of cell density followed the same pattern as for @-galactosidaseproduction for the two cell lines. The sensitivity of the T n 5B1-4 cells to cell density agreed with earlier published results showing an effect of cell density on NOV production from T n 368 cells (Wood et al., 1982). Further, these authors showed that the cell contact inhibition resulted in cessation of viral DNA replication. They also observed that when medium from infected cells grown at high density (no polyhedra produced) was used with infected cells at low density, polyhedra and NOV were produced, indicating that the buildup of a toxic metabolite or depletion of a nutrient was not responsible for the cell density effect. For the other cell lines, the density that gave peak @galactosidase production was used to measure the titer of infectious virus produced per cell as shown in Table 11.A comparison of the cell lines on a per cell basis showed up to 3.5-fold differences in NOV titers with Ea 88 showing the highest production. However, when NOV production was normalized on a per microgram of uninfected cell
The advantage of the recombinant protein used in this study is its ease of assay which allows the quantitative evaluation of many cell lines under different conditions. E. coli @-galactosidasehas been used for this reason in a number of studies involving the baculovirus expression system (Pennock et al., 1984; King et al., 1991; Licari and Bailey, 1991);however, a major drawback of this protein is that it is not secreted and it undergoes no posttranslational modifications-many proteins of commercial interest undergo some form of posttranslational modifications. A majority of the proteins being expressed using the baculovirus expression system are secreted, glycosylated proteins whose expression levels are much lower than those of E. coli @-galactosidaseor polyhedrin (Luckow, 1990). Davis et al(1992a) have recently developed a baculovirus vector expressing human placental secretory alkaline phosphatase (seAP) which is glycosylated, secreted, and easily assayed using colorimetric techniques similar to those used for @-galactosidase. This vector has also been used to screen insect cell lines that are superior to Sf 9 (Davis et al., 1992b). Cell contact inhibition is not normally an important factor to consider when cells are grown in suspension. However, the importance of cell density inhibition is revealed when cell lines are analyzed for their production in conditions other than suspension-such as attached to wells. All the cell lines except for Sf 9 and Sf 21 exhibited moderate to severe cell density inhibition. Wood et al. (1982) found that Tn 368 cell-cell contact between logphase cells plated at high cell densities caused a dramatic reduction in polyhedrin production and NOV titers, which was directly correlated with a cessation of viral DNA replication. Inhibition by nutrient depletion or a buildup of toxic byproducts was not found to be a major factor contributing to the decreases in polyhedra and NOV production in cells at low density using conditioned medium from infected cells at high density. In the present study, all the cell lines were infected at low density during logarithmic growth and then were plated in fresh medium. Significant inhibition by nutrient depletion or buildup of toxic byproducts was unlikely because a 4-fold range of medium volumes (halving and doubling the normal volume) while keeping the cells per plate constant had little effect on the specific cellular @-galactosidaseproduction. While the presence of oxygen limitations cannot be completely ruled out by changing medium volumes, it was found that similar reductions in 8-galactosidase production as a function of cell densities occurred when
395
Biotechnol. Prog., 1992, Vol. 8, No. 5
T n 5B1-4cells were grown as attached monolayers in roller bottles, where oxygen limitations due to overlyingmedium cannot occur. Furthermore, the effect of density on @galactosidase production was time-dependent such that cells infected at low density and then plated at high density 24 h pi produced 2.5- and 4-fold higher @-galactosidase levels than cells plated to high density at 12 and 0 h pi (data not shown), respectively. Such large changes in production are more likely to be caused by an event as overriding of the normal cellular processes (i.e., inhibition of DNA synthesis by cell-cell contact) during viral replication than by nutrient limitations or toxic product inhibition that would occur in the 12-24-h period. The effect of cell density on virus production from T n 5B1-4 cells was similar to the effect observed in T n 368 cells by Wood et al. (1982). Virus production from T n 5B1-4and Sf 9 cells followed the same patterns of inhibition with cell density as those for @-galactosidaseproduction. However, no correlation was found between a cell line's @-galactosidaseproduction and its NOV production. King et al. (1991) found that an unspecified M. brassicae cell line, which produced at least twice as much @-galactosidase as Sf 21 cells, produced 20-fold lower titers of NOV. This finding suggested an inverse relation might exist between recombinant protein and NOV production. However, Tables I and I1 show that while the cell line Mb 0507 produced more @-galactosidaseper milliliter as Sf 21, similar to the levels reported by King et al. (1991), there was little variation in NOV production for seven of the cell lines that were tested. Hink et al. (1991) also tested 23 cell lines for the production of @-galactosidaseand two glycosylated, secreted proteins. Both their study and the present study tested @-galactosidaseproduction from the cell lines IPLB Sf 21AE, Sf 9, and T n 368. However, direct comparisons are difficult for two reasons. First, the @-galactosidase vectors used in the two studies were different, which presumably resulted in different expression levels. The highest expression level obtained by Hink et al. (1991) at 3 days, converted to international units, was 22.6 IU/106 Sf 9 cells. By comparison, the level of @-galactosidase produced from Sf 9 cells in the present study was 112 IU/106 cells. This difference is unlikely to be due to cell density effects given the insensitivity of Sf 9 cells to this factor. A second difficulty in comparing the relative differences among cell lines is the differences in the percentage of infected cells achieved for the same cell lines between the two studies. Comparison of the same cell lines that have been passaged separately is difficult since different populations are most likely selected during passaging. For example, over 90% of the Sf 21 and T n 368 cells used in this study became productively infected compared to the 4% and 17% infections, respectively, observed by Hink et al. (1991). These differences could reflect the selection of cells that invariably must occur during passaging. For example, the cell line BTI EaA (Granados and Naughton, 1976) has been reported to achieve only 19% infection; however, after passaging for two years, the cell line Ea 88 was derived from these cells which is over 85 3' 6 susceptible (Granados et al., manuscript in preparation). The effect of passaging on a cell line is thus an important factor to consider in comparing protein production levels reported for the same cell lines. A major question that remains to be resolved for the insect cell/baculovirus system is why many posttranslationally-modified proteins do not achieve the expression levels of polyhedrin or @-galactosidase. Secreted, glycosylated proteins represent a major portion of recombinant
molecules that are being produced using the baculovirus expression vector (Possee et al., 1990; Luckow, 1990). Unfortunately, over 100-fold reductions in the amount of glycosylated, secreted proteins synthesized compared to polyhedrin and @-galactosidaseare routinely observed. Certainly, vector construction plays a major role; however, for many vectors the mRNA expression levels of polyhedrin and many recombinant genes are nearly the same in Sf 9 cells (Luckow, 19901,suggesting that some other factor also plays a strong role in the reductions in protein expression. Jarvis and Summers (1989) found that addition of the oligosaccharide structure to the protein was rapid, on the basis of the predominance of the glycosylated form of tissue plasminogen activator (TPA) over the nonglycosylated form during pulsechase experiments in Sf 9 cells. However, they also found that Sf 9 cells had a 1.6-h half-time of secretion for TPA and suggested that secretion of TPA was rate-limited at a step or steps following translocation and N-linked glycosylation. Observations by Vernet et al. (1990) also suggested that the secretion of propapain, which requires proteolytic cleavage and glycosylation for activity, was limited by a step(s) early in the secretory pathway and by step(@following N-linked glycosylation in Sf 9 cells. Other cell lines may exist which are less rate-limited than Sf 9 cells. The T n 5B1-4 cell line displays a number of characteristics that make it desirable for use in the baculovirus expression system. It produced 7 times higher 8-galactosidase per cell and 2.2 times higher @-galactosidaseper milliliter than Sf 9 cells. The 5B1-4 cells produce over 25 times more secreted, glycosylated alkaline phosphatase than Sf 9 cells (Davis et al., 1992; Wickham, 1991). For two other secreted, glycosylated proteins T n 5B1-4 cells have been found to produce 5-28-fold more recombinant protein per cell that Sf 9 cells in serum-free medium (Wickham and Nemerow, 1992). The major disadvantage of the T n 5B1-4 cell line, at present, is that ita growth is attachment-dependent. However, many insect cell lines that have grown attached have been adapted to suspension cultures. Additionally, T n 5B1-4 cells have been grown in a packed, glass-bead reactor (Shuler et al, 1990). Cell contact inhibition can be also avoided by providing a high surface area within the reactor while at the same time achieving cell densities comparable to Sf 9 cells in free suspension culture. The screeningprocedure reported here to identify cell lines superior to Sf 9 in @-galactosidaseor secretory alkaline phosphatase (Davis et al., 1992b) production can be used to rapidly identify other promising new cell lines for their production of a secreted (alkaline phosphatase) or nonsecreted (@-galactosidase)protein using the colorimetric assays that are available for each protein.
Acknowledgment This work was supported by the National Science Foundation, who sponsored this research under (NSF EET-8807089),and by the Cornel1BiotechnologyProgram in the form of a pre-doctoral fellowship awarded to T.J.W. The Cornel1 Biotechnology Program is sponsored by the New York State Science and Technology Foundation, a consortium of industries, the U S . Army Research Office, and the National Science Foundation.
Literature Cited Davis, T.; Munkenbeck-Trotter,K.; Granados, R. R.; Wood, H. A. Expression of secreted mammalian alkaline phosphatase with baculovirus. J . Gen. Virol.,submitted for publication, 1992a.
Biotechnol. Prog., 1992, Vol. 8, No. 5
398
Davis, T. R.; Wickham, T. J.; Granados, R. R.; Wood, H. A. Comparative recombinant protein expression in eight insect cell lines. In Vitro Mol. Cell. Biol., submitted for publication, 1992b.
Goodwin,R. H. Insect cellculture: improvedmedia and methods for initiating attached cell lines from the lepidoptera. In Vitro 1975,11, 369-378.
Granados, R. R.;Naughton, M. Replication of Amsacta moorei entomoposvirus in ovarian and hemocyte cultures from Estigmene acrea larvae. In Invertebrate Tissue Culture,Applications in Medicine, Biology, and Agriculture; Kutstak, F., Maramorasch, K., Eds.; Academic Press: New York, 1976; pp 472-476. Granados, R. R.; Derksen, C. G.; Dwiyer,K. G. Replication of the Trichoplusia ni granulosis and nuclear polyhedrosis viruses in cell cultures. Virology 1986, 152, 472-476. Greenfield, C.; Patel, G.; Clark, S.; Jones, N.; Waterfield, M. D. Expression of the human EGF receptor with ligand-stimulatable kinase activity in insect cells using a baculovirusvector. EMBO J . 1988, 7, 139-146. Hink, W. F. Established insect cell line from the cabbage looper, Trichoplusia ni. Nature (London) 1970,266,466-467. Hink, W. F.; Thomsen, D. R.; Davidson, D. J.; Meyer, A. L.; Castellino,F. J. Expressionof three recombinant proteins using baculovirus vectors in 23 insect cell lines. Biotechnol. Prog. 1991, 7, 9-14.
Jarvis, D. L.; Summers, M. D. Glycosylation and secretion of human tissue plasminogen activator in recombinant baculovirus-infectedinsect cells. Mol. Cell. Biol. 1989,9,214-223. King, L. A.; Mann, S. G.; Laorie, A. M.; Mulshaw, S. H. Replication of wild-type and recombinant Autographa californica nuclear polyhedrosisvirus in a cell line derived form Mamestra brassicae. Virus Res. 1991, 19, 93-104.
Licari, P.; Bailey,J. E. Factors influencing recombinant protein yields in an insect cell-baculovirusexpression system: multiplicity of infection and intracellular protein degradation. Biotechnol. Bioeng. 1991,37,238-246.
Luckow, V. A. Cloning and expression of heterologous genes in insect cells with baculovirus vectors. In Recombinant DNA Technology and Applications; Ho, C., Prokop, A., Baipai, R., Eds.; McGraw-Hill: New York, 1990; pp 1-24. Moore, M. D.; Cannon, M. J.; Sewall,A.; Finlayson,M.; Okimoto, M.; Nemorow, G. R. Inhibition of Epstein-Barrvirus infection in vitro and in vivo by soluble CR2 (CD21) containing two short consensus repeats. J. Virol. 1991, 65, 3559-3565. Pennock,G. D.;Shoemaker, C.; Miller,L. K. Strongandregulated expression of Escherichia coli &galactosidase in insect cell with a baculovirus vector. Mol. Cell. Biol. 1984, 4, 399-406. Shuler, M. L.; Cho, T.; Wickham, T. J.; Ogonah, 0.; Kool, M.; Hammer, D. A.; Granados, R. R.; Wood, H. A. Bioreactor development for production of viral pesticides or heterologous
proteins in insect cell cultures. Ann. N.Y. Acad. Sci. 1990, 589,399-422.
Smith, G. E.; Jr, G.; Ericson, B. L.; Moschera, J.; Lahm, H.-W.; Chizzonite,R.; Summers, M. D. Modification and secretion of human interleukin 2 produced in insect cells by a baculovirus expressionvector. Proc. Natl. Acad. Sci. U.S.A.1985,82,84048408.
Steiner, H.; Poh., G.; Gunne, H.; Hellers, M.; Elhammer, A.; Hansson, L. Human tissue-type plasminogen activator synthesized by using a baculovirus vector in insect cells compared with human plasminogen activator produced in mouse cells. Gene 1988, 73,449-457.
Summers, M. D.; Smith, G. A. A manual of methods for baculovirus vectors and insect cell culture procedures. Ten. Agric. Exp. Stn. Bull. 1987, B1555, 1-56.
Vaughn, J. L.; Goodwin, R. H.; Tompkins, G. J.; McCawley, P. The establishment of two cell linesfrom the insect Spodoptera frugiperda (Lepid0ptera:Noctuidae).In Vitro 1977,13,213217.
Vernet, T.; Tessier, D. C.; Richardson, C.; Laliberte, F.; Khouri, H. E.; Bell, A. W.; Storer, A. C.; Thomas, D. Y. Secretion of functional papain precursor from insect cells. J. Biol. Chem. 1990,265, 16661-16666.
Whang, Y.; Silberklang, M.; Morgan, A.; Munshi, S.; Lenny, A. B.; Ellis, R. W.; Kieff, E. Expression of the Epstein-Barrvirus gp350/220gene in rodent and primate cells. J . Virol. 1987,61, 1796-1807.
Wickham,T. J. Baculovirus-insectcell interactions in producing heterologous proteins: attachment, infection, and expression in different cell lines. Ph.D. Disseration, Cornel1University, Ithaca, NY, 1991. Wickham,T. J.; Nemerow,G. R. Optimization of growthmethods and recombinant protein production in BTI-Tn-5B1-4 insect cells using the baculovirus expression system. Biotechnol. Prog., submitted for publication, 1992. Wojchowski,D. M.; Orkin, St. H.; Sytkowski,A. J. Active human erhthropoietin expressed in insect cells using a baculovirus vector: a role for N-linked oligosaccharide. Biochim. Biophys. Acta 1987, 910, 224-232.
Wood, H. A. An agar overlayplaque assay method for Autographa californica nuclear polyhedrosis virus. J . Inuertebr. Pathol. 1977,29, 304-307.
Wood, H. A. Autographa californica nuclear polyhedrosisvirusinduced proteins in tissue culture. Virology 1980,102,21-27. Wood, H. A.; Hohnston, L. B.; Burand, J. P. Inhibition of Autographa californica nuclear polyhedrosisvirus replication in high-density Trichoplusia ni cell cultures. Virology 1982, 119, 245.
Accepted June 21, 1992. Registry No. @-Galactosidase,9031-11-2.
