Professional Documents
Culture Documents
DOI 10.1002/btpr.238
Published online July 27, 2009 in Wiley InterScience (www.interscience.wiley.com).
A chemically defined nutrient feed (CDF) coupled with basal medium preloading was
developed to replace a hydrolysate-containing feed (HCF) for a fed-batch NS0 process. The
CDF not only enabled a completely chemically defined process but also increased recombi-
nant monoclonal antibody titer by 115%. Subsequent tests of CDF in a CHO process indi-
cated that it could also replace the hydrolysate-containing nutrient feed in this expression
system as well as providing an 80% increase in product titer. In both CDF NS0 and CHO
processes, the peak lactate concentrations were lower and, more interestingly, lactate metab-
olism shifted markedly from net production to net consumption when cells transitioned from
exponential to stationary growth phase. Subsequent investigations of the lactate metabolic
shift in the CHO CDF process were carried out to identify the cause(s) of the metabolic
shift. These investigations revealed several metabolic features of the CHO cell line that we
studied. First, glucose consumption and lactate consumption are strictly complementary to
each other. The combined cell specific glucose and lactate consumption rate was a constant
across exponential and stationary growth phases. Second, Lactate dehydrogenase (LDH) ac-
tivity fluctuated during the fed-batch process. LDH activity was at the lowest when lactate
concentration started to decrease. Third, a steep cross plasma membrane glucose gradient
exists. Intracellular glucose concentration was more than two orders of magnitude lower
than that in the medium. Fourth, a large quantity of citrate was diverted out of mitochondria
to the medium, suggesting a partially truncated tricarboxylic acid (TCA) cycle in CHO cells.
Finally, other intermediates in or linked to the glycolytic pathway and the TCA cycle, which
include alanine, citrate, isocitrate, and succinate, demonstrated a metabolic shift similar to
that of lactate. Interestingly, all these metabolites are either in or linked to the pathway
downstream of pyruvate, but upstream of fumarate in glucose metabolism. Although the spe-
cific mechanisms for the metabolic shift of lactate and other metabolites remain to be eluci-
dated, the increased understanding of the metabolism of CHO cultures could lead to future
improvements in medium and process development. V C 2009 American Institute of Chemical
Introduction the cells or are required at much lower levels than supplied.
Third, the chemically undefined nature of hydrolysates hin-
Hydrolysates, especially those of nonanimal origin, are ders the scientists’ understanding process of cell metabolism.
widely used as serum replacements in therapeutic protein For instance, it is not well understood how mammalian cells
production processes using mammalian cells. Hydrolysates metabolize amino acids in peptides, so that amino acid con-
can improve cell growth and protein yield in serum-free sumption analysis based on free amino acids measurements
processes (Burteau et al., 2003; Heidemann et al., 2000; may or may not be accurate (Nyberg et al., 1999).
Sung et al., 2004; Xie et al., 1997). However, there are a
few limitations that complicate hydrolysates’ utilization. First Development of various chemically defined media for
of all, lot-to-lot variation exists for hydrolysates (Luo and NS0 cells and CHO cells has been reported (Chen et al.,
Chen, 2007; Zhang et al., 2003). Prescreening using small- 2000; Hata et al., 1992; Huang et al., 2007; Zhang and Rob-
scale performance testing is typically required to eliminate inson, 2005), and commercial chemically defined media are
underperforming lots. Second, the nutrient composition in readily available from all major mammalian cell culture me-
hydrolysates is not balanced for cell consumption. Various dium manufacturers. However, the development of chemi-
components, such as ash and salts, are either not required by cally defined fed-batch processes that support rapid cell
proliferation, high cell density, and high production rate has
been challenging. Chemically defined fed-batch processes
Correspondence concerning this article should be addressed to N. Ma have been reported recently. Gong et al. (2006) developed a
at ningning.ma@pfizer.com. chemically defined fed-batch process for hybridoma cells.
Peak viable cell density reached about 8 106 cells/mL, but tate metabolic shift from net production to net consumption.
viable cells dropped sharply after a 5-day culture. Burky At the end of the process, lactate concentration in the CDF
et al. (2006) developed a generic chemically defined fed- processes was almost undetectable. Subsequent investigations
batch process for NS0 cells. The process supported peak via- were carried out to better understand this metabolic shift in
ble cell density around 1 107 cells/mL and product titer of the CDF CHO process.
2.64 g/L on Day 13.
Lactic acid, or lactate, has long been recognized as a Materials and Methods
major by-product in cell culture processes (Hu et al., 1987;
Miller et al., 1989; Ozturk and Palsson, 1992). Base addition Cells
to neutralize lactic acid elevates medium osmolarity, which, NS0 and CHO cells producing test recombinant monoclo-
in turn, inhibits cell growth and causes viability to drop. nal antibodies were used in this study. Both cell types used
Reducing lactic acid production had been shown to increase glutamine synthatase as the transfection selection marker.
peak cell density, extend process duration, and increase
recombinant protein yield. Four strategies had been reported
in the literature to reduce lactate production. The first one is Process
to reduce medium glucose concentration to limit glucose NS0 and CHO cells were routinely subcultured in 250 mL
supply (Cruz et al., 1999; Zhou et al., 1995, 1997b). Using shake flasks (Corning, Acton, MA) in chemically defined hy-
an online dynamic feeding protocol, Zhou et al. (1995, bridoma and CHO media, respectively. Both hybridoma and
1997b) successfully maintained glucose at a low level of CHO media are commercially available. The CHO medium was
about 0.5 mM. As a result, lactate production was reduced supplemented with 25 mM methionine sulfoximine (MSX).
and higher peak cell densities and higher viability were Two-liter Applikon bioreactors (Applikon, Foster city,
achieved. A second strategy is to use alternative, slow-con- CA) with a 1-L working volume were used in this study. In
suming carbon sources (Altamirano et al., 2004). Altamirano these bioreactors, two pitched blade impellers provided mixing
et al. (2004) used a feeding strategy in which glucose and and gas bubble dispersion. Temperature was controlled at
galactose feeds were alternated. This strategy forced CHO 36.5 C using a heating blanket. pH was controlled at 7.0 by
cells to consume more residual lactate than the culture fed carbon dioxide sparging through a ring sparger and by adding
with glucose only. As a result, harvest viability was mod- either 0.5 N sodium hydroxide or 7.5% sodium bicarbonate.
estly improved. The third strategy is to downregulate lactate Dissolved oxygen (DO) was controlled at 30% air saturation
dehydrogenase-A (LDH-A) (the LDH subunit that effectively using air overlay and oxygen sparging. The fed-batch process
catalyzes pyruvate to lactate conversion) using a wide vari- for NS0 and CHO cells was very similar. Cells are seeded at
ety of techniques including homologous recombination 2.5 105 viable cells/mL in corresponding NS0 or CHO pro-
(Chen et al., 2001), antisense mRNA (Jeong et al., 2001; duction media. Starting from Day 3, a nutrient feed was added
Paredes et al., 1999), and small interfering RNA (Kim and into the bioreactors at 28 mL/L/day and 16 mL/L/day for NS0
Lee, 2007a). Chen et al. (2001) compared a LDH-A downre- and CHO cells, respectively, till the end of the process. Glucose
gulated hybridoma clone to a control clone. In the culture was fed separately to control glucose concentration around 2 g/
utilizing the LDH-A downregulated cell line, lactate concen- L. Bioreactors were sampled daily for offline measurements.
tration was about 35% lower and monoclonal antibody titer
was about 200% higher. However, although Kim and Lee
(2007a) reduced lactate production rate even further (by 45– Routine bioreactors offline measurements
79%), the recombinant thrombopoietin product rates of three Cell density and viability were measured using a Cedex
LDH-A downregulated clones were not statistically better cell counter (Innovatis, Bielefeld, Germany). Offline DO,
than the control clone. A fourth strategy is to enhance the pH, glucose concentration, and lactate concentration were
flow of glucose carbon into tricarboxylic acid (TCA) cycle. measured with a BioProfile chemical analyzer (Nova Bio-
A focus area has been the overexpression of pyruvate car- medical, Waltham, MA). A portion of the broth from each
boxylase of yeast (Irani et al., 1999, 2002) or human origin daily sampling was filtered through a 0.2 lm polyethersul-
(Kim and Lee, 2007b). In a perfusion culture, Irani et al. fone filter (Pall, Ease Hills, NY) for product titer and amino
(2002) demonstrated that pyruvate carboxylase-transfected acid measurements.
BHK-21 cells produced half of the lactate when compared Monoclonal antibody titer was measured using an Agilent
with the control cells. Cell density was about 50% higher HPLC (Hewlett Packard, Waldbronn, Germany) with a
and recombinant erythropoietin production rate was about POROS protein-A affinity column (Applied Biosystems, Fos-
100% higher. More recently, it was reported that the expres- ter City, CA). The column was calibrated using correspond-
sion of antiapoptosis genes, E1B-19K and Aven (EA167), ing reference standards before titer quantification.
could reduce lactate production of CHO cells (Dorai et al.,
in press). In batch cultures with or without glucose limita-
tion, a CHO clone transfected with antiapoptotic genes gave Intracellular LDH activity measurement
lower lactate concentration compared with the control cell About 2–4 mL of broth was taken out of bioreactors and
line at corresponding conditions. transferred to a centrifuge tube chilled on ice. The cells were
In this study, we developed a chemically defined nutrient washed twice with cold phosphate buffered saline (PBS). In
feed (CDF) with companion production medium preloading each wash, the cells were pelletted in a refrigerated centri-
to replace a hydrolysate-containing feed (HCF) for a fed- fuge at 150g for 5 min, the supernatant was discarded, and
batch NS0 process. The CDF was later investigated in a the cell pellet was gently resuspended in 4 mL of cold PBS.
CHO process. In both NS0 and CHO processes, CDF After the second wash, cells were diluted to 5 105 viable
resulted in higher product titer and lower peak lactate con- cells/mL and lysed using freeze-thaw. In the freeze-thaw
centration. The CDF processes also exhibited a dramatic lac- procedure, cells were first flash frozen in liquid nitrogen and
Biotechnol. Prog., 2009, Vol. 25, No. 5 1355
Metabolic shift of glycolytic pathway and TCA link closely to the central metabolic pathway, to understand
cycle intermediates the mechanism(s).
Metabolites discussed in the previous sections are sum-
marized in Figure 10. The intermediates showing a transition Conclusions
from net accumulation to net consumption are in bold, and
the intermediates that were undetectable or not monitored A CDF with companion medium preloading for a fed-
are labeled with an asterisk. It is interesting to note that the batch NS0 process was successfully developed to replace
shifted metabolites are clustered together, downstream of py- a HCF. The new feed not only allowed us to achieve a
ruvate and upstream of fumarate with citrate considered the fully chemically defined process but also offered higher
start of the TCA cycle. Of all the intermediates monitored in peak cell density and higher product titer. When the CDF
this segment, a-ketoglutarate is the only one showing a dif- with its companion medium preloading was tested in a
ferent metabolic profile from lactate. The similar metabolism CHO process, it again demonstrated multiple significant
of a group of closely related metabolites suggests that the benefits. In addition to allowing the removal of hydroly-
metabolic shift was controlled by flux or by a group of sates, both viability and product titer were substantially
highly coordinated enzymes. A possible flux controlling improved. A lower lactate concentration is an additional
point is the synthesis of pyruvate from phosphoenolpyruvate. benefit of the CDF processes in both NS0 and CHO cul-
Again, the flux limitation hypothesis does not agree with the tures. In the CDF processes, there was a rapid shift in lac-
increased glucose consumption rate (Figure 4b from Stage II tate metabolism that resulted in a nearly complete
to Stage III). More studies are needed to elucidate the true depletion of lactate in the second half of the process. It
mechanism of the metabolic shift. First, in this study, we was also observed that the lactate metabolic shift occurred
monitored only a single enzyme (LDH). A more broad pro- concurrently with the transition of cell growth from expo-
teomics study is needed to assist interpreting the metabolo- nential to stationary phase.
mics results. Second, the metabolic analysis was largely Analysis of net glucose consumption by CHO cells indi-
based on trend analysis. A quantitative flux analysis could cated that the glycolytic pathway and pyruvate to lactate
increase the quality of the analysis and might lead to new conversion are closely coordinated. Their flows are strictly
findings. Third, it might be necessary to consider other path- complementary to each other so that the combined cell spe-
ways, such as urea cycle and lipid synthesis pathways that cific glucose and lactate consumption rate is a constant. A
1362 Biotechnol. Prog., 2009, Vol. 25, No. 5
plausible factor that could exert this tight control is cellular Literature Cited
maintenance of the NADþ/NADH balance. Altamirano C, Paredes C, Illanes A, Cairo JJ, Godia F. Strategies
LDH activity oscillated during the process. The lowest ac- for fed-batch cultivation of t-PA producing CHO cells: substitu-
tivity measured was during the lactate metabolic shift from tion of glucose and glutamine and rational design of culture me-
net accumulation to net consumption. However, the results dium. J Biotechnol. 2004;110:171–179.
Baggetto LG. Deviant energetic metabolism of glycolytic cancer
seem to be counterintuitive so that a more in-depth study of cells. Biochimie. 1992;74:959–974.
the LDH subunits and the other pair of substrates in the lac- Board M, Hummt S, Newsholme EA. Maximum activities of key
tate–pyruvate reaction, NADþ and NADH, is needed to enzymes of glycolysis, glutaminolysis, pentose phosphate pathway
determine the link between the LDH activity change and the and tricarboxylic acid cycle in normal, neoplastic and suppressed
lactate metabolic shift. cells. Biochem J. 1990;265:503–509.
Bogan JS, McKee AE, Lodish HF. Insulin-responsive compartments
Additional studies with the CHO process indicated that containing GLUT4 in 3T3-L1 and CHO cells: regulation by
several amino acids demonstrated metabolic shifts as well. amino acid concentrations. Mol Cell Biol. 2001;21:4785–4806.
Alanine, another byproduct of pyruvate, showed a similar Burky JE, Wesson MC, Young A, Farnsworth S, Dionne B, Zhu Y,
net accumulation to net consumption shift as lactate. This Hartman TE, Qu L, Zhou W, Sauer PW. Protein-free fed-batch
demonstrates that the metabolic shift is not limited to pyru- culture of non-GS NS0 cell lines for production of recombinant
vate–lactate conversion only. antibodies. Biotechnol Bioeng. 2006;96:281–293.
Burteau CC, Verhoeye FR, Mols J, Ballez JS, Agathos SN, Schnei-
Finally, profiling of most glycolytic and TCA cycle inter- der YJ. Fortification of a protein-free cell culture medium with
mediates in the CHO CDF process was conducted. In this plant peptones improves cultivation and productivity of an inter-
study, both intracellular and extracellular metabolite levels feron-c -producing CHO cell line. In Vitro Cell Dev Biol Anim.
were measured. A steep glucose concentration gradient was 2003;39:291–296.
observed across the plasma membrane, where intracellular Chen K, Liu Q, Xie L, Sharp PA, Wang DIC. Engineering of a mam-
malian cell line for reduction of lactate formation and high mono-
glucose concentration was more than two orders of magni-
clonal antibody production. Biotechnol and Bioeng. 2001;72:55–61.
tude lower than the extracellular concentration. The low glu- Chen Z, Iding K, Lukemeyer D, Lehmann. A low-cost chemically
cose concentration is likely due to the high phosphorylation defined medium for a recombinant CHO cell line producing pro-
rate of glucose in the cytosol, rather than a glucose transpor- thrombin. Biotechnol Lett. 2000;22:837–841.
tation limitation. Cruz HJ, Moreira JL, Carrondo MJT. Metabolic shifts by nutrient
manipulation in continuous cultures of BHK cells. Biotechnol and
When lactate was nearly depleted, the medium pyruvate
Bioeng. 1999;66:104–113.
became undetectable in the CHO CDF process. Meanwhile, Dang CV, Semenza GL. Oncogenic alterations of metabolism.
two glycolytic intermediates upstream of pyruvate formation, 3- Trends Biochem Sci. 1999;24:68–72.
phosphoglycerate and phosphoenolpyruvate, accumulated in the Dorai H, Kyung YS, Ellis D, Kinny C, Lin C, Jan D, Moore G,
medium. The opposite profile between phosphoenolpyruvate and Betenbaugh MJ. Expression of anti-apoptotic genes alters lactate
pyruvate suggests that the flux from the former to the latter metabolism of Chinese hamster ovary cells in culture. Biotechnol
might be partially blocked around the time at which the lactate Bioeng. In press.
Godia F, Cairo JJ. Cell metabolism. In:Ozturk SS,Hu W-S, editor-
metabolic shift occurred. However, this hypothesis is not sup- s.Cell Culture Technology for Pharmaceutical and Cell-Based
ported by the glucose consumption data, where glucose con- Therapies. New York: Taylor and Francis; 2006. pp 81–112.
sumption rate increased after the lactate metabolic shift. Gong X, Li D, Li X, Fang Q, Han X, Wu Y, Yang S, Shen BQ.
A considerable amount of citrate was exported out of mi- Fed-batch culture optimization of a growth-associated hybridoma
tochondria into the medium in the first 9-day culture, after cell line in chemically defined protein-free media. Cytotechnol-
ogy. 2006;52:25–38.
which citrate metabolism converted from net accumulation Harrison SA, Buxton JM, Czech MP. Suppressed intrinsic catalytic
to net consumption in the CHO CDF process. The extracel- activity of GLUT1 glucose transporters in insulin-sensitive 3T3-
lular citrate concentration increased up to 10-fold higher L1 adipocytes. Proc Natl Acad Sci USA. 1991;88:7839–7843.
than the intracellular average concentration. This observation Hata J, Tamura T, Yokoshima S, Yamashita S, Kabeno S, Matsu-
suggests that the TCA cycle in CHO cells might be partially moto K, Onadera K. Chemically defined medium for the
truncated. Besides citrate, isocitrate and succinate also dem- production of biologically active substances of CHO cells. Cyto-
technology. 1992;10:9–14.
onstrated similar metabolic shift as lactate and alanine. More Hatzivassiliou G, Zhao F, Bauer DE, Andreadis C, Shaw AN, Dhanak D,
interestingly, all the metabolites showing a metabolic shift Hingorani SR, Tuveson DA, Thompson CB. ATP citrate lyase inhibi-
are clustered together, generally downstream of pyruvate and tion can suppress tumor cell growth. Cancer Cell. 2005;8:311–321.
upstream of fumarate. Heidemann R, Zhang C, Qi H, Rule JL, Rozales C, Park S, Chuppa
Although the CDF fed-batch process demonstrated signifi- S, Ray M, Michaels J, Konstantinov K, Naveh D. The use of pep-
tones as medium additives for the production of a recombinant
cant empirical benefits over the previous HCF process and therapeutic protein in high density perfusion cultures of mamma-
provides a platform for the study of cellular metabolism, the lian cells. Cytotechnology. 2000;32:157–167.
cause(s) of the metabolic shifts in the CDF CHO process is Hu WS, Dodge TC, Frame KK, Himes VB. Effect of glucose on
still unknown. It could be a flux restriction, where a possible the cultivation of mammalian cells. Developments in biological
restriction point is the conversion of phosphoenolpyruvate to standardization. 1987;66:279–290.
pyruvate or a systematic enzymatic shift during cell growth Huang EP, Marquis CP, Gray PP. Development of super-CHO pro-
tein-free medium based on a statistical design. J Chem Technol
phase transition. Follow-up studies, especially a companion
Biotechnol. 2007;82:431–441.
proteomics study, will help elucidate the mechanism(s). Irani N, Beccaria AJ, Wagner R. Expression of recombinant cyto-
plasmic yeast pyruvate carboxylase for the improvement of the
production of human erythropoietin by recombinant BHK-21
Acknowledgments cells. J Biotechnol. 2002;93:269–282.
Irani N, Wirth M, van Den Heuvel J, Wagner R. Improvement of
The authors appreciate Dr. Kurt Droms and Dr. Amit Bane- the primary metabolism of cell cultures by introducing a new
rjee for their critical review of the manuscript and Process De- cytoplasmic pyruvate carboxylase reaction. Biotechnol and Bio-
velopment Analytical group for the amino acid analyses. eng. 1999;66:238–246.
Biotechnol. Prog., 2009, Vol. 25, No. 5 1363
Jeong DW, Kim TS, Lee JW, Kim KT, Kim HJ, Kim IH, Kim IY. Peterson PL, Mathupala S, Rempel A, Geschwind JF, Ko YH. Mito-
Blocking of acidosis-mediated apoptosis by a reduction of lactate chondrial bound type II kinase: a key player in the growth and
dehydrogenase activity through antisense mRNA expression. Bio- survival of many cancers and an ideal prospect for therapeutic
chem Biophys Res Commun. 2001;289:1141–1149. intervention. Biochim Biophys Acta. 2002;1555:14–20.
Kim SH, Lee GM. Down-regulation of lactate dehydrogenase-A by Sung YH, Lim SW, Chung JY, Lee GM. Yeast hydrolysate as a
siRNAs for reduced lactic acid formation of Chinese hamster low-cost additive to serum-free medium for the production of
ovary cells producing thrombopoietin. Appl Microbiol Biotechnol. human thrombopoietin in suspension cultures of Chinese hamster
2007a;74:152–159. ovary cells. Appl Microbiol Biotechnol. 2004;63:527–536.
Kim SH, Lee GM. Functional expression of human pyruvate carbox- Tsao YS, Cardoso AG, Condon RGG, Voloch M, Lio P, Lagos JC,
ylase for reduced lactic acid formation of Chinese hamster ovary Kearns BG, Liu Z. Monitoring Chinese hamster ovary cell culture
cells (DG44). Appl Microbiol Biotechnol. 2007b;76:659–665. by the analysis of glucose and lactate metabolism. J Biotechnol.
Kondoh H, Lleonart ME, Gil J, Wang J, Degan P, Peters G, Marti- 2005;118:316–327.
nez D, Carnero A, Beach D. Glycolytic enzymes can modulate Warburg O.The Metabolism of Tumors. London: Arnold Constable;
cellular life span. Cancer Res. 2005;65:177–185. 1930.
Lawton KA, Berger A, Mitchell M, Milgram KE, Evans AM, Guo Xie L, Nyberg G, Gu X, Li H, Möllborn F, Wang DIC. Gammain-
L, Hanson RW, Kalhan SC, Ryals JA, Milburn MV. Pharmacoge- terferon production and quality in stoichiometric fed-batch cul-
nomics. 2008;9:383–397. tures of Chinese hamster ovary (CHO) cells under serum-free
Luo Y, Chen G. Combined approach of NMR and chemometrics for conditions. Biotechnol Bioeng. 1997;56:577–582.
screening peptones used in the cell culture medium for the pro- \Xie L, Wang DIC. Stoichiometric analysis of animal cell growth and its
duction of a recombinant therapeutic protein. Biotechnol Bioeng. application in medium design. Biotechnol Bioeng. 1994;43:1164–1174.
2007;97:1654–1659. Zhang J, Reddy J, Buckland B, Greasham R. Toward consistent and
Macheda ML, Rogers S, Best JD. Molecular and cellular regulation productive complex media for industrial fermentations: studies on
of glucose transporter (GLUT) proteins in cancer. J Cell Physiol. yeast extract for a recombinant yeast fermentation process. Bio-
2005;202:654–662. technol Bioeng. 2003;82:640–652.
Miller WM, Wilke CR, Blanch HW. Transient responses of hy- Zhang J, Robinson D. Development of animal-free, protein-free, and
bridoma cells to nutrient additions in continuous culture: I. Glu- chemically-defined media for NS0 cell culture. Cytotechnology.
cose pulse and step changes. Biotechnol and Bioeng. 1989;33: 2005;48:59–74.
477–486. Zhou W, Chen CC, Buckland B, Aunins J. Fed-batch culture of
Nyberg GB, Balcarcel RR, Follstad BD, Stephanopoulos G, Wang Recombinant NS0 myeloma cells with high monoclonal antibody
DIC. Metabolism of peptide amino acids by Chinese hamster production. Biotechnol Bioeng. 1997a;55:783–792.
ovary cells grown in a complex medium. Biotechnol Bioeng. Zhou W, Rehm J, Hu WS. High viable cell concentration fed-batch
1999;62:324–335. cultures of hybridoma cells through online nutrient feeding. Bio-
Ozturk SS, Riley MR, Palsson BO. Effects of ammonia and lactate technol and Bioeng. 1995;46:579–587.
on hybridoma growth, metabolism, and antibody production. Bio- Zhou W, Rehm J, Europa A, Hu WS. Alteration of mammalian cell
technol and Bioeng. 1992;39:418–431. metabolism by dynamic nutrient feeding. Cytotechnology. 1997b;
Paredes C, Prats E, Cairos JJ, Azorin F, Cornudella Ll, Godia F. 24:99–108.
Modification of glucose and glutamate metabolism in hybridoma
cells through metabolic engineering. Cytotechnology. 1999;30:
85–93. Manuscript received Sept. 15, 2008, and revision received Feb. 28, 2009.