Reduction of CMP-N-acetylneuraminic acid hydroxylase activity in engineered Chinese hamster ovary cells using an antisense-RNA strategy Stephane Chenu a , Anne Gre ´goire a , Yanina Malykh b , Athanase Visvikis c , Lucia Monaco d,1 , Lee Shaw b , Roland Schauer b , Annie Marc a , Jean-Louis Goergen a, * a Laboratoire des Sciences du Ge ´nie Chimique, CNRS-ENSAIA, 2, av. de la Fore ˆt de Haye, F-54505 Vandoeuvre-le `s-Nancy, France b Biochemisches Institut, Christian-Albrechts-Universita ¨t zu Kiel, Olshausenstr. 40, D-24098 Kiel, Germany c Faculte ´ de Pharmacie, Univ. H. Poincare ´, 30 rue Lionnois, F-54000 Nancy, France d DIBIT – San Raffaele Scientific Institute, I-20132 Milan, Italy Received 2 April 2003; received in revised form 5 June 2003; accepted 18 June 2003 Abstract Rodent cells, widely used for the industrial production of recombinant human glycoproteins, possess CMP-N-acetylneuraminic acid hydroxylase (CMP-Neu5Ac hydroxylase; EC 1.14.13.45) which is the key enzyme in the formation of the sialic acid, N-glycolylneuraminic acid (Neu5Gc). This enzyme is not expressed in an active form in man and evidence suggests that the presence of Neu5Gc in recombinant therapeutic glycoproteins may elicit an immune response. The aim of this work was, therefore, to reduce CMP-Neu5Ac hydroxylase activity in a Chinese Hamster Ovary (CHO) cell line, and thus the Neu5Gc content of the resulting glycoconjugates, using a rational antisense RNA approach. For this purpose, the cDNA of the hamster hydroxylase was partially cloned and sequenced. Based on the sequence of the mouse and hamster cDNAs, optimal antisense RNA fragments were selected from preliminary in vitro translation tests. Compared to the parental cell line, the new strain (CHO-AsUH2), which was transfected with a 199-bp antisense fragment derived from the mouse CMP-Neu5Ac hydroxylase cDNA, showed an 80% reduction in hydroxylase activity. An analysis of the sialic acids present in the cells’ own glycoconjugates revealed a decrease in the percentage of Neu5Gc residues from 4% in the parental cells to less than 1% in the CHO-AsUH2 cell line. D 2003 Elsevier B.V. All rights reserved. Keywords: CMP-Neu5Ac hydroxylase; Sialylation; N-glycolylneuraminic acid; CHO cell; Antisense RNA 1. Introduction Chinese Hamster Ovary (CHO) cells are widely used for the production of recombinant glycoproteins employed in human therapy. Although the posttranslational modifica- tions of recombinant glycoproteins generated in CHO cells are very similar to that present on human proteins, these rodent cells do not perfectly mimic the human sialylation pattern required for bioactivity, stability, and absence of antigenicity [1,2]. One of the most significant differences between rodent and human sialylation is the incorporation of N-glycolylneuraminic acid (Neu5Gc). This sialic acid is present in the glycoconjugates of virtually all vertebrates, with the notable exception of man and chicken. In animals, Neu5Gc is formed as the sugar-nucleotide cytidine mono- phosphate-N-glycolylneuraminic acid (CMP-Neu5Gc) by the action of CMP-Neu5Ac hydroxylase (EC 1.14.13.45) [3]. In man, the gene encoding this enzyme lacks an exon, resulting in an incomplete mRNA with a premature termi- nation signal [4,5]. Due to this mutation, Neu5Gc cannot be formed by CMP-Neu5Ac hydroxylase in man and is prac- tically undetectable in healthy human tissues [6], though small amounts of Neu5Gc are found in gangliosides and glycoproteins of certain human tumors [7]. Glycoconjugates containing this common sialic acid are therefore foreign to humans and can induce an immune response, causing the formation of so-called Hanganutziu– Deicher antibodies [8,9]. The appearance of these ‘‘serum sickness’’ antibodies was observed after inoculation of patients with animal 0304-4165/03/$ - see front matter D 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0304-4165(03)00137-5 * Corresponding author. Tel.: +33-383-59-5844; fax: +33-383-59- 5804. E-mail address: [email protected](J.-L. Goergen). 1 Present address: Keryos SpA, via Maritano 26, I-20097 San Donato Milanese, Italy. www.bba-direct.com Biochimica et Biophysica Acta 1622 (2003) 133 – 144
12
Embed
Reduction of CMP N-acetylneuraminic acid hydroxylase activity in engineered Chinese hamster ovary cells using an antisense-RNA strategy
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
www.bba-direct.com
Biochimica et Biophysica Acta 1622 (2003) 133–144
Reduction of CMP-N-acetylneuraminic acid hydroxylase activity in
engineered Chinese hamster ovary cells using an antisense-RNA strategy
Stephane Chenua, Anne Gregoirea, Yanina Malykhb, Athanase Visvikisc, Lucia Monacod,1,Lee Shawb, Roland Schauerb, Annie Marca, Jean-Louis Goergena,*
aLaboratoire des Sciences du Genie Chimique, CNRS-ENSAIA, 2, av. de la Foret de Haye, F-54505 Vandoeuvre-les-Nancy, FrancebBiochemisches Institut, Christian-Albrechts-Universitat zu Kiel, Olshausenstr. 40, D-24098 Kiel, Germany
cFaculte de Pharmacie, Univ. H. Poincare, 30 rue Lionnois, F-54000 Nancy, FrancedDIBIT–San Raffaele Scientific Institute, I-20132 Milan, Italy
Received 2 April 2003; received in revised form 5 June 2003; accepted 18 June 2003
Abstract
Rodent cells, widely used for the industrial production of recombinant human glycoproteins, possess CMP-N-acetylneuraminic acid
hydroxylase (CMP-Neu5Ac hydroxylase; EC 1.14.13.45) which is the key enzyme in the formation of the sialic acid, N-glycolylneuraminic
acid (Neu5Gc). This enzyme is not expressed in an active form in man and evidence suggests that the presence of Neu5Gc in recombinant
therapeutic glycoproteins may elicit an immune response. The aim of this work was, therefore, to reduce CMP-Neu5Ac hydroxylase activity
in a Chinese Hamster Ovary (CHO) cell line, and thus the Neu5Gc content of the resulting glycoconjugates, using a rational antisense RNA
approach. For this purpose, the cDNA of the hamster hydroxylase was partially cloned and sequenced. Based on the sequence of the mouse
and hamster cDNAs, optimal antisense RNA fragments were selected from preliminary in vitro translation tests. Compared to the parental
cell line, the new strain (CHO-AsUH2), which was transfected with a 199-bp antisense fragment derived from the mouse CMP-Neu5Ac
hydroxylase cDNA, showed an 80% reduction in hydroxylase activity. An analysis of the sialic acids present in the cells’ own
glycoconjugates revealed a decrease in the percentage of Neu5Gc residues from 4% in the parental cells to less than 1% in the CHO-AsUH2
cell line.
D 2003 Elsevier B.V. All rights reserved.
Keywords: CMP-Neu5Ac hydroxylase; Sialylation; N-glycolylneuraminic acid; CHO cell; Antisense RNA
1. Introduction of N-glycolylneuraminic acid (Neu5Gc). This sialic acid is
Chinese Hamster Ovary (CHO) cells are widely used for
the production of recombinant glycoproteins employed in
human therapy. Although the posttranslational modifica-
tions of recombinant glycoproteins generated in CHO cells
are very similar to that present on human proteins, these
rodent cells do not perfectly mimic the human sialylation
pattern required for bioactivity, stability, and absence of
antigenicity [1,2]. One of the most significant differences
between rodent and human sialylation is the incorporation
0304-4165/03/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
was used for gene amplification at a starting concentration
of 50 nM and was increased progressively to a final
concentration of 500 nM. Clones were isolated after a 9-
week period in the selection medium. For sialic acid
analyses, cells were maintained in the same selection
medium except that calf serum was replaced by 4% human
serum.
2.5. Preparation of high-speed supernatants from different
cell lines for CMP-Neu5Ac hydroxylase assays
Cell pellets (about 108 cells) were thawed, resuspended
in one volume (ml/g wet mass) of ice-cold 50 mM HEPES/
NaOH, pH 7.4 containing Triton X-100 (5% by mass) and
incubated for 20 min at 4 jC with permanent agitation. The
resulting homogenates were centrifuged for 15 min at
100000� g and 4 jC and the clear high-speed supernatants
were removed and used for determination of the hydroxy-
lase activity.
2.6. Assay of CMP-Neu5Ac hydroxylase activity
The activity of CMP-Neu5Ac hydroxylase was detected
by incubation of 15 Al of the high-speed supernatants with 1
mM NADH, 0.5 mM FeSO4, pig liver microsomes (60 Agprotein solubilised in a final concentration of 0.5% w/v
Triton X-100), and CMP-[4,5,6,7,8,9-14C]Neu5Ac (Amer-
sham) in 50 mM HEPES/NaOH, pH 7.4 at 37 jC, the finalvolume of the assay being 25 Al. The concentration of CMP-
Neu5Ac in the test varied from 2.1 to 18 AM and the
quantity of the radioactivity was 12.5–15.6 nCi per assay.
To obtain substrate concentrations above 12 AM the re-
quired amount of non-radioactive CMP-Neu5Ac was added.
Incubation times are as indicated in Results and all enzyme
tests were performed in duplicate. The assays were stopped
by the addition of 5 Al 1 M trichloroacetic acid, and
precipitated proteins were removed by centrifugation for 3
min at 14000� g. Released [14C]Neu5Gc and [14C]Neu5Ac
were separated and quantified by radio thin-layer chroma-
tography, as described previously [31]. Minute amounts of
activity due to traces of hydroxylase in the pig microsomes,
were subtracted from the sample activity. With this method,
the minimum detectable conversion of CMP-Neu5Ac to
CMP-Neu5Gc was about 1–2%. The activity of each
sample was estimated using two incubation times, both of
which were within the linear region of the reaction time
course (see Results). The apparent KM for CMP-Neu5Ac of
the enzyme in the extracts of CHO-UH cells was determined
using the non-linear regression program Enzfitter (Elsevier
Biosoft, UK).
2.7. RNA analysis by RT-PCR
Total RNA was prepared from CHO-UH and CHO-
AsUH2 cells (SV Total RNA isolation system, Promega)
and the first cDNA strand synthesis was performed using
the AMV reverse transcriptase and random nonamers.
Oligonucleotides specific for the hamster CMP-Neu5Ac
hydroxylase and the expression vector pCISfiT sequences
were used to amplify either the CMP-Neu5Ac hydroxylase
cDNAs or the 199 bp fragments present in the cell lines.
PCR amplification was performed in a final volume of 50
Al, using 0.8 units of AccuTaq polymerase (Sigma) and 1 Agof template cDNA. Amplification was carried out using 35
cycles with a temperature profile of: 40 s at 94 jC, 1 min at
58 jC and 1.5 min at 68 jC. The PCR products were
analysed by electrophoresis on a 1.5% agarose gel. A
similar protocol was used to amplify a h-actin sequence,
with the help of standard oligonucleotides, except that the
temperature of the annealing step was increased to 68 jC.
2.8. Sialic acid analysis
Cells (6–10� 107) were thawed, resuspended in two
volumes (ml/g wet mass) of distilled water and homoge-
nised on ice for 3� 10 s using a Branson B-12 sonicator (40
W power) fitted with a microtip, allowing 15 s cooling
between bursts. To release sialic acids, the resulting homog-
enate (0.1 ml) was mixed with the same volume of 0.2 M
HCl and incubated for 2 h at 80 jC. The samples were
centrifuged for 20 min at 14000� g and the resulting pellets
were washed with 0.1 ml ice-cold water followed by
centrifugation at 14000� g. After both centrifugation steps,
the supernatants were pooled and filtered with a Centricon
10 ultrafiltration unit (Amicon). The retained material was
washed twice with 0.2 ml water followed by ultrafiltration.
The filtrates were lyophilised, and the released sialic acids
S. Chenu et al. / Biochimica et Biophysica Acta 1622 (2003) 133–144 137
were purified by passage through 2 ml cation- and anion-
exchange columns as described by Ref. [32]. The adsorbed
sialic acids were eluted from the latter resin with seven
column volumes of 1 M formic acid, lyophilised and
resuspended in 1 ml water. The enriched sialic acids were
derivatized with 1,2-diamino-4,5-methylene dioxybenzene
(DMB) and the resulting derivatives were analysed by
reversed-phase HPLC with fluorimetric detection [33].
Quantification of sialic acids was performed by integration
of the peaks in the chromatogram using a calibration curve
obtained by DMB-derivatization of known Neu5Ac
amounts.
2.9. Protein concentration
Protein concentrations were measured using the BCA
protein assay reagent (Pierce) with bovine serum albumin as
a standard.
3. Results
3.1. PCR amplification and cloning of the partial CMP-
Neu5Ac hydroxylase cDNA from CHO-cells
Using primers designed from the nucleotide sequence of
the mouse CMP-Neu5Ac hydroxylase cDNA [19], two
products of 1691 and 1250 bp were amplified by RT-PCR
using total RNA from CHO-UH cells as a template (data not
shown). These fragments were subcloned into pCR2.1
TOPO and transformed into E. coli. Three colonies from
each transformation were randomly selected, their plasmids
isolated and the DNA inserts sequenced. The amino acid
sequence deduced from the ORF of the 1250-bp product
exhibited a 98% similarity to fibronectin from CHO, a
protein which is abundantly expressed in CHO cells. The
nucleotide sequence of the 1691-bp product was 94%
identical to that of the mouse CMP-Neu5Ac hydroxylase
cDNA. The deduced sequence of the 563 amino acids was
97% homologous to the mouse enzyme, indicating that
Fig. 2. Schematic representation of the mouse CMP-Neu5Ac hydroxylase cDNA cl
translation inhibition tests.
several nucleotide exchanges are silent (Fig. 1). Due to
the fact that oligonucleotides from the mouse hydroxylase
open reading frame were used to amplify the PCR frag-
ments, the given sequence is only partial, covering 98% of
the mouse CMP-Neu5Ac hydroxylase cDNA. The high
degree of homology between the sequence presented here
and the primary structure of CMP-Neu5Ac hydroxylase
from mouse and other sources [34] confirms its assignment
as CMP-Neu5Ac hydroxylase, presumably the main en-
zyme responsible for the biosynthesis of Neu5Gc in CHO
cells.
3.2. Selection of CMP-Neu5Ac hydroxylase antisense
constructs
Since initial experiments [35,36] showed that artificial
antisense oligonucleotides may be used to interfere with
gene expression, there has been a growing body of literature
on antisense nucleic acids [22,23,37–39]. Due to the lack of
generally applicable rules for the design of efficient anti-
sense RNAs, the ability of different antisense RNAs to
inhibit the translation of the mouse CMP-Neu5Ac hydrox-
ylase mRNA was tested in vitro. The approach adopted
involved examining the efficiency of antisense RNAs of
different lengths by generating fragments from the mouse
CMP-Neu5Ac hydroxylase cDNA. Four different antisense
RNAs of 714, 403, and 199 nucleotides (nt) plus the full-
length of the mouse cDNA were generated (for exact
position on the hydroxylase cDNA see materials and meth-
ods). In addition, a 199 nt fragment from the hamster CMP-
Neu5Ac hydroxylase cDNA spanning the same region as
the mouse 199 nt fragment was tested (Fig. 2).
The mouse CMP-Neu5Ac hydroxylase cDNA and the
antisense fragments were transcribed in vitro with either T7
or T3 RNA polymerases and the formation of the respective
RNA was confirmed by agarose gel electrophoresis by
comparison with four standard RNAs of 250, 1065, 1525,
and 2346 nt (data not shown). The band intensity reflected
the amount of RNA obtained, allowing an estimation of the
sense and antisense RNA quantities used in the in vitro
oned into pGEM3Z and of the sense and antisense RNAs used in the in vitro
Fig. 4. pCISfiT plasmid used for the cloning and the in vitro amplification
of the antisense RNAs. pCISfiT possesses a multiple cloning site terminated
by a SfiI cloning site; the gene of interest is under the control of the strong
CMV promoter; a SV40 late polyadenylation sequence and a transcription
terminator from the human gastrin gene are located downstream to the
cDNA coding for the antisense.
Fig. 3. In vitro CMP-Neu5Ac hydroxylase translation inhibition tests. Proteins translated using rabbit reticulocyte lysates were loaded on a SDS
polyacrylamide gel, then transferred on an Immobilon P membrane. Incorporation of biotinylated lysine during translation was detected as described in the
‘‘Transcendk non-radioactive translation detection’’ kit. Panels A and B: antisense/sense RNAs molar ratio = 1:1. Panels C and D: antisense/sense RNAs
molar ratio = 1:2. For all panels, T(– ): proteins naturally biotinylated in the reticulocyte lysate. Luciferase: positive control resulting from the in vitro luciferase
mRNA translation. CMP-Neu5Ac hydroxylase: in vitro translation of the CMP-Neu5Ac hydroxylase mRNA; Full-length, 714mAs, 403mAs, 199mAs and
199hAs: in vitro translation of the CMP-Neu5Ac hydroxylase mRNA in the presence of the mentioned antisense RNA fragments (numbers show the length of
fragments; m and h indicate the mouse or hamster source of the hydroxylase cDNA, used to design the fragments; arrows indicate the position of the CMP-
Neu5Ac hydroxylase).
S. Chenu et al. / Biochimica et Biophysica Acta 1622 (2003) 133–144138
translation tests. Fig. 3A and B present the proteins trans-
lated in vitro with different levels of sense and antisense
RNAs. Luciferase is the positive control giving rise to a
band of 61 kDa. As can be seen, when the same amount of
sense and antisense RNA were mixed in the translation
cocktail, no band corresponding to the CMP-Neu5Ac hy-
droxylase protein (66 kDa) appeared on SDS-polyacryl-
amide gel electrophoresis of the products, indicating that
the five antisense RNAs were able to inhibit the translation
of CMP-Neu5Ac hydroxylase mRNA. After a 50% reduc-
tion in the quantity of the antisense RNAs, the full-length,
the 403 and the 199 nucleotides still led to a very efficient
inhibition of translation, whereas a strong reduction of the
inhibition was observed with the 714 nt antisense RNA (Fig.
3C and D). Interestingly, both the mouse and hamster 199
bp antisense fragments were equally effective in inhibiting
hydroxylase translation, despite 15 nt differences. In view of
these results, the shorter and most efficient antisense frag-
S. Chenu et al. / Biochimica et Biophysica Acta 1622 (2003) 133–144 139
ments were selected and cloned into eucaryotic expression
vectors for the in vivo inhibition of CMP-Neu5Ac hydrox-
ylase translation.
3.3. Generation of CMP-Neu5Ac hydroxylase depleted cell
lines
EcoRI–XbaI antisense fragments of 199 and 403 bp
were cloned into the pCISfiT vector (Fig. 4) and expressed
under the control of the CMV promoter. Each of these
constructs was combined with the pSfiSVdhfr plasmid,
which bears the DHFR selectable marker, according to the
in vitro amplification method [30], for the insertion of
multiple copies of the antisense fragment into the genome
Fig. 5. Radio thin-layer chromatograms of [14C] sialic acids released by acid trea
lines with 2.1 AM CMP-[14C]Neu5Ac and cofactors, as described in Materials and
Al of high-speed supernatant of OKT3 mouse hybridoma cells. Panel B (negative c
supernatant. Panels C and D: analysis of the parental CHO-UH cell line for 60 and
and 403m2 cell lines, respectively. The numbers above the peaks identify the nat
of CHO-UH cells. DHFR-positive transformants were
obtained after 9 weeks by initial selection in aMEM
medium lacking ribo- and deoxyribonucleosides, followed
by exposure to progressively increasing concentrations of
MTX up to 500 nM. One of the main advantages of the in
vitro amplification system is the high rate of good producer
clones generated and then a reduction of the clones to be
screened. The combined in vitro/in vivo amplification
procedures was already used for the production of granulo-
cyte colony-stimulating factor (G-CSF) [29]: a 30-fold
improvement of the protein production was obtained com-
pared to the in vitro amplification protocol alone.
A total of approximately 15 clones was obtained, of
which two to three stable clones from each of the three
tment of incubation mixtures of high-speed supernatants of the various cell
methods. The time of incubation is indicated. Panel A (positive control): 15
ontrol): 15 Al of the 50 mM HEPES/NaOH was used instead of high-speed
120 min, respectively. Panels E and F: analysis for 120 min of the 199m2
ure of the sialic acid ((1) Neu5Gc; (2) Neu5Ac).
S. Chenu et al. / Biochimica et Biophysica Acta 1622 (2003) 133–144140
antisense constructs were tested for their CMP-Neu5Ac
hydroxylase activity.
3.4. Assay of CMP-Neu5Ac hydroxylase activity in
engineered cell lines
Representative radio thin-layer chromatograms of the
enzyme tests for several cell lines are given in Fig. 5. The
activity of the enzyme in the parental CHO-UH cells was
very low in comparison with other cell lines, such as OKT3
mouse hybridoma cells which are known to produce large
amounts of Neu5Gc (Fig. 5A and C). To obtain a high
turnover of the radioactive substrate and thus the required
sensitivity, it was necessary to use non-isotopically diluted
CMP-[14C]Neu5Ac in low concentrations with long incu-
bation times. For example, after a 2-h incubation, the high-
speed supernatant of CHO-UH cells in the assays containing
2.1 and 18 AM substrate gave rise to 14% and 3%
[14C]Neu5Gc, respectively.
The apparent KM for CMP-Neu5Ac determined for the
crude extracts of CHO-UH cells was about 8 AM (data not
shown); a comparable value for this parameter had been
obtained for purified mouse (5.5 AM) and porcine (11 AM)
hydroxylases [40,41]. However, a substrate concentration of
2.1 AM, which allowed the reliable determination of
[14C]Neu5Gc, was used to compare hydroxylase activities
of the various CHO clones. This CMP-Neu5Ac concentra-
tion was significantly lower than KM. Nevertheless, sub-
strate consumption was always less than 15% (Fig. 5C–F)
and increased linearly with the incubation time (Fig. 5C and
D) suggesting that the above results allow a quantitative
comparison of the enzyme activity of the various clones
presented in Fig. 6. To confirm this, the activity of the
Fig. 6. Specific activity of the CMP-Neu5Ac hydroxylase in different CHO-clones,
four separate measurements is shown. Standard deviations are denoted with error ba
8 measurements.
hydroxylase in these cell lines was also measured with
substrate concentrations of 12 and 18 AM, both of which
are greater than KM (data not shown). The less accurate data
obtained for the substrate concentrations above the KM
revealed the same tendency as in Fig. 6 (data not shown).
A reduction of the hydroxylase activity in comparison
with parental UH cells was observed in seven out of the
eight clones studied (Fig. 6). The greatest reduction of
hydroxylase activity was observed in the cell line 199m2,
in which a 78% reduction of the activity measured in the
parental CHO-UH cells was detected (Figs. 4E and 5). This
antisense cell line was therefore chosen for further inves-
tigations and renamed CHO-AsUH2.
3.5. mRNA expression profile
The mRNA expression pattern of CMP-Neu5Ac hydrox-
ylase was studied by RT-PCR analysis on both CHO-UH
parental and the selected CHO-AsUH2 cells (Fig. 7). Total
RNA was isolated from both cell lines and reversed tran-
scribed using random nonamers. Two different 35 cycle-
PCR reactions were then performed with two sets of
primers, using the same quantity of the respective template
cDNA. The first PCR served to amplify a portion of the
hydroxylase cDNA, whereas the second was aimed at the
amplification of the 199 nt antisense RNA. As shown in Fig.
7B, the presence of the 199 nt antisense RNA fragment was
confirmed in the CHO-AsUH2 cell line, and as expected it
was absent in the parental cell line. In order to compare the
level of hydroxylase mRNA expression between both cell
lines, the relative amounts of the amplified RT-PCR prod-
ucts were quantified using three different gels, by densito-
metric analysis on each (Fig. 7A). Although the band
determined using 2.1 AM CMP-[14C]Neu5Ac. The mean specific activity of
rs. For the definition of clones, see legend to Fig. 4. (*) Data for 12 and (**)
Fig. 8. Sialic acid composition of glycoconjugates from CHO-UH or CHO-
AsUH2 cells. DMB-modified sialic acids were separated by reversed phase
HPLC coupled to fluorimetric detection. Upper panel: sialic acids extracted
from glycoconjugates of CHO-UH cells. Lower panel: sialic acids from
glycoconjugates of CHO-AsUH2 cells.
S. Chenu et al. / Biochimica et Biophysica Acta 1622 (2003) 133–144 141
corresponding to the hydroxylase mRNA was present in
both cell lines, an 80% (F 6%) decrease in its intensity was
observed in the CHO-AsUH2 cells. As a control for RNA
quality and to assess RT-PCR conditions, single-stranded
cDNA was used as a template for the amplification of h-actin with specific forward and reverse primers. In all
experiments, the h-actin specific primers amplified the
expected 550 bp product (data not shown). The formation
of the hydroxylase PCR-product in CHO-UH cells had not
reached a plateau, since a 40 cycle amplification yielded a
band of higher intensity.
3.6. Analysis of the sialic acids in CHO-UH and CHO-
AsUH2 cell lines
For the sialic acid analyses, both parental and CHO-
AsUH2 cells were cultured for 32 passages in medium
supplemented with 4% human serum, which is devoid of
Neu5Gc residues. This culture medium containing human
serum (not relevant from a biotechnological point of view
for the production of a recombinant protein) was selected
with the aim of removing all traces of Neu5Gc, which could
appear in the cells as the result of the uptake and utilization
of the Neu5Gc-containing glycoconjugates from FCS. How-
ever, the type of serum added to the culture medium had no
influence on the activity of the hydroxylase, i.e. the specific
enzyme activities were 190F 25 fmol/min/mg protein for
CHO-UH cells and 42F 10 fmol/min/mg protein for CHO-
AsUH2 cells grown in presence of calf or human serum.
The total amount of sialic acid was practically the same
in both cell lines grown in FCS-free medium (CHO-UH
cells: 1.8 Ag sialic acid/mg protein; CHO-AsUH2 cells: 1.5
Ag sialic acid/mg protein). Significantly, the amount of
Fig. 7. RT-PCR analysis of CHO-UH and CHO-AsUH2 cell lines. The PCR
products were loaded on a 1.5% agarose gel. Panel A: CMP-Neu5Ac
hydroxylase mRNA amplification in CHO-UH and CHO-AsUH2 cell lines,
respectively. Panel B: 199m-antisense RNA amplification in CHO-UH and
CHO-AsUH2 cell lines, respectively. For both panels, MW=DNA
molecular weight marker.
Neu5Gc as a percentage of total sialic acids was consider-
ably decreased in the CHO-AsUH2 cells (0.62F 0.05%)
compared with the CHO-UH cells (3.7F 0.3%) (Fig. 8).
4. Discussion
The aim of this work was to reduce CMP-Neu5Ac
hydroxylase activity in a CHO cell line, and thus generate
a stable cell line performing a glycosylation as similar as
possible to that present in humans. This involved obtaining
cells which produce a2,6-linked sialic acids and lacked
Neu5Gc. To achieve this, a genetically modified ‘‘universal
host’’ cell line CHO-UH [2], stably expressing a2,6 sialyl-
transferase, was used in an antisense RNA approach to
inhibit the translation of CMP-Neu5Ac hydroxylase.
Despite a growing body of literature on the use of
antisense nucleic acids in modifying gene expression
[22,23,35–39], no general rules for the selection of an
efficient antisense RNA have yet emerged. In order to
design an optimal antisense RNA fragment, partial nucleo-
tide sequence of the hamster CMP-Neu5Ac hydroxylase
was cloned by RT-PCR amplification. The deduced nucle-
otide sequence of the hamster hydroxylase (Fig. 1) was
highly homologous to that of the mouse enzyme and the
hydroxylase from other mammals [5,34].
Various antisense RNAs of 714, 403, and 199 nucleo-
tides, derived from the mouse and hamster sequences, and a
full-length antisense RNA from the mouse cDNA (Fig. 3)
S. Chenu et al. / Biochimica et Biophysica Acta 1622 (2003) 133–144142
were preliminary tested in hydroxylase translation inhibition
in vitro tests with different amounts of antisense and sense
RNAs. All antisense RNAs, independent of the animal
source, were able to inhibit the translation of the hydroxy-
lase in vitro, though the 714 nt fragment was markedly less
effective, probably due to the formation of stable secondary
structures. The shorter antisense RNAs, which appeared to
be most efficient, were therefore used to generate stable
transfectants of the CHO-UH cells.
Several approaches were tested to analyse which of the
cell lines expressing different antisense fragments were most
efficient at suppressing hydroxylase production. However,
due to the very low level of enzyme in the parental CHO-
UH cells, the sensitivity of the detection methods was
generally insufficient. For example, the screening of a
CHO cDNA library using the full-length mouse CMP-
Neu5Ac hydroxylase cDNA as a probe, failed to detect
any hydroxylase-specific clones (data not shown). Accord-
ingly, a RT-PCR analysis showed that mRNA coding for
CMP-Neu5Ac hydroxylase is expressed at a very low level
in CHO-UH cells (Fig. 7A). Furthermore, a sensitive
Western blot analysis with a CMP-Neu5Ac hydroxylase-
specific antibody, at a detection limit of about 20 ng
enzyme, failed to reveal the hydroxylase in 90 Ag of total
protein from high-speed supernatants of CHO-UH cells
resolved by SDS-PAGE (data not shown). Determining
Neu5Gc was also a misleading measure of cellular hydrox-
ylase, due to the small amount of this sialic acid (3.7% of
total sialic acid) in the parental CHO-UH cell line. This
latter analysis is further complicated by the fact that cells in
culture incorporate Neu5Gc from FCS present in the medi-
um [42,43]. For this reason, screening the clones by sialic
acid analysis was only meaningful after cultivating cells for
a long period of time in FCS-free medium. Despite the fact
that sub-KM CMP-[14C]Neu5Ac concentrations had to be
used, the sensitive radio-TLC-based assay provided a quan-
titative measure of the relative hydroxylase content of the
various cell lines.
Using this enzyme assay, it was first shown that most of
the tested clones displayed a reduced CMP-Neu5Ac hy-
droxylase activity; in fact, this is probably due to the use of
the combined in vitro/in vivo amplification procedures
yielding a high rate of strong producer clones. Second, the
efficiency of the antisense fragments in vivo was indepen-
dent of the animal source, and the 199 nt fragments were
more effective than the 403 nt antisense RNAs at suppress-
ing hydroxylase expression (Fig. 6). The cell line trans-
fected with the 199 nt fragment derived from the cDNA of
the mouse enzyme exhibited the greatest (78%) reduction in
the hydroxylase activity and was termed CHO-AsUH2.
Although the latter antisense fragment differed from the
corresponding hamster fragment by 15 nt, both were effi-
cient at inhibiting the translation of the hydroxylase mRNA
in vivo. In fact, these sequence differences could be advan-
tageous, since natural antisense RNAs do not exhibit com-
plete complementary to the target mRNA and thus are less
prone to intracellular RNases [44,45]. Two different mech-
anisms could account for the inhibition of translation by
antisense fragments [37]: (i) steric blocking of translation
initiation, and (ii) degradation of the antisense RNA-mRNA
duplex by specific RNases. However, since none of the 199
nt antisense fragments was located upstream the AUG
initiator codon, the reduction of CMP-Neu5Ac hydroxylase
activity in CHO-AsUH2 is probably mainly due to the
degradation of mRNAs. Indeed, the semi-quantitative esti-
mation of the amounts of mRNA by RT-PCR showed a
significant reduction (approximately 80%) in CMP-Neu5Ac
hydroxylase mRNA in the CHO-AsUH2 cell line as com-
pared to CHO-UH cells (Fig. 7A). This result is also in a
good agreement with previous findings demonstrating that
(antisense RNA-sense mRNA) hybrids are rapidly degraded
by RNases specific for RNA duplexes, leading to a decrease
in the endogenous message levels of a targeted gene [46,47].
An analysis of Neu5Gc, the product of the hydroxylase
reaction, in total cellular glycoconjugates produced by cells
cultivated for 32 passages in the presence of human serum,
revealed an 80% decrease in the amount of Neu5Gc in the
CHO-AsUH2 cell line relative to the parental cells (Fig. 8).
Since a reduction in CMP-Neu5Ac hydroxylase activity of
about 80% led to the same decrease in the amount of
Neu5Gc residues present on the glycoconjugates in the
new CHO-AsUH2 cell line, the amount of hydroxylase
protein appears to be governing the incorporation of
Neu5Gc into glycoconjugates. This is in agreement with
previous studies on various mammalian tissues [33,41,48] in
which the amount of Neu5Gc was found to be regulated by
the level of hydroxylase activity. This, in turn, generally
correlates with the quantity of hydroxylase protein. More-
over, the estimated 80% reduction in the levels of hydrox-
ylase mRNA found in CHO-AsUH2 cells suggests that in
this cell line, the observed decrease in the CMP-Neu5Ac
hydroxylase activity and the amount of Neu5Gc are due to
the presence of antisense RNA.
Although we were able to significantly reduce the CMP-
Neu5Ac hydroxylase activity in CHO-UH cells, the anti-
sense RNA approach does not yield a complete suppression
of the enzyme activity. This agrees with previous reports on
the inhibition of sialidase activity in CHO cells [23]. Despite
the fact that the CMP-Neu5Ac hydroxylase mRNA only
seems to be present in a low copy number, and the antisense
fragment was amplified by means of DHFR-MTX selection
and in vitro concatenamer amplification systems, the CHO-
AsUH2 cells still exhibited 20% of the parental enzyme
activity. Nevertheless, compared to previous studies [21,23,
38], the approach described here enabled the generation of a
stable cell line which retained a significant reduction of the
target enzyme activity for more than 30 passages. Though
the overproduction of a recombinant protein by this cell line
might require the use of a selection/amplification marker
different from DHFR/MTX (for instance glutamine syn-
thase/methionine sulfoximine) [49], these new CHO-
AsUH2 cells are suitable for industrial purposes and further
S. Chenu et al. / Biochimica et Biophysica Acta 1622 (2003) 133–144 143
studies will focus on the assessment of the reduction of
Neu5Gc levels in recombinant glycoproteins produced by
these cells.
Acknowledgements
This work was partially supported by the European
community, grant No. 960767. The authors thank Dr.
Kawano for providing the pBSCNAH vector containing the
mouse CMP-Neu5Ac hydroxylase cDNA.
References
[1] N. Jenkins, E.M. Curling, Glycosylation of recombinant proteins: prob-
lems and prospects. Enzyme Microb. Technol. 16 (1994) 354–364.
[2] A. Bragonzi, G. Distefano, L.D. Buckberry, G. Acerbis, C. Foglieni,
D. Lamotte, G. Campi, A. Marc, M.R. Soria, N. Jenkins, L. Monaco,
A new Chinese hamster ovary cell line expressing a2,6-sialyltransfer-
ase used as universal host for the production of human-like sialylated