the most expressed construct for each plant type studied. We also calculated the probability of the
values obtained being significantly different using the Student law. We considered as significantly
different those pairs of values 5 or less likely to share a same group.
3. Results
3
.
1
. Nucleotide frequencies of the translation initiation context in different tissue types
The starting point for this work was the identifi- cation of small open reading frames within the
leader of several transcripts encoding a plasma membrane H
+
-ATPase [25,30]. We obtained evi- dence that these open reading frames were trans-
lated by a fraction of the scanning ribosomes, although they did not have an AUG context close
to the consensus. This prompted us to examine the AUG context of plant genes. We therefore re-
trieved the AUG context from the genes available in the databases. Using the 5075 rule introduced
by Caverner [4], we deduced the consensus con- texts as a A AC a AUG G C and c AG CA C
AUG G C for dicot and monocot plants, respec- tively. As these data were largely confirmed by a
recent study [11], we will not comment further on them.
3
.
2
. Selection of AUG contexts for experimental e6aluation
To experimentally evaluate the importance of the AUG context, we chose the firefly luciferase
luc gene [24] as a reporter. The sequence sur- rounding the initiation AUG was replaced with a
cassette allowing an exchange of 16 different AUG contexts Table 1. As a negative control, we used
a construct with the deleted AUG initiator codon construct 17. Besides the comparison of monocot
and dicot consensus contexts, we considered the importance of the residues at positions − 3, − 2,
−
1 and + 4, + 5 using the dicot consensus as a starting point Table 1. We also calculated the
frequency of each of the 16 contexts in both dicot and monocot databases Table 1. It appears that
the consensus contexts derived for dicots are in- deed the most frequent of the 16 contexts used in
this study. Concerning the monocot database, two contexts constructs 6 and 7 that are frequent in
dicot genes, are as frequent or more frequent than the consensus sequences derived for monocots.
This indicates that the consensus context, which is a juxtaposition of the most frequent nucleotides
at each position, is not necessarily the most fre- quent context. The 16 sequences tested here corre-
spond to 13.9 of the genes in the dicot database and 9.3 of the genes in the monocot database.
Although in vitro translation systems have been used to study the cell translation machinery, they
might not reflect the real conditions and organisa- tion of the translation system [21,31] and might
therefore introduce some bias in the expression of the test constructs.
Transgenic plants, in which the test gene is integrated in the genome, would not be convenient
for a quantitative comparison between several constructs because the so-called ‘position effect’
would introduce expression variations that might be much larger than the variations associated with
the AUG context. In this case, RNA quantifica- tion would also be required, thus introducing a
further parameter. We therefore relied on a tran- sient expression system, using either electropora-
tion or bombardment. As an internal control, we used a plasmid carrying the Renilla luciferase [32].
The reporter and control luc genes were both placed under the control of the cauliflower mosaic
virus 35S promoter.
3
.
3
. Assay of the AUG context in tobacco leaf protoplasts
In tobacco protoplasts, the highest expression level was obtained with construct 1, which had a
consensus sequence for dicots Fig. 1A, construct 1. A control without AUG had no activity con-
struct 17. The second dicot consensus context C instead of A at position − 2 had a LUC activity
significantly see Section 2 for the definition of this word reduced by approximately 30 construct
6. Expression of constructs 2, 3 and 4 revealed that substitution at position − 3 of A by any
other nucleotide also significantly reduced LUC activity by about one third. The importance of A,
and not just as a purine A or G as it sometimes appears in consensus sequences [7] at position
−
3, was also supported by the comparison of constructs 1 with 2, 6 with 7 and 12 with 13.
A similar significant reduction in expression re- sulted from the substitution of A by C at − 2 or
− 1 positions constructs 6 and 5 versus 1. This
negative effect was not cumulative when both − 1 and − 2 positions were substituted by C con-
struct 12. Changing four nucleotides − 4 to − 1 up-
stream from the AUG anti-upstream consensus, construct 10 reduced LUC activity more than the
substitution of two nucleotides + 4, + 5 down- stream from the AUG anti-downstream consen-
sus, construct 11. Combining anti-upstream and anti-downstream consensus sequences resulted in a
further significant LUC reduction down to 10 construct 9. The two monocot consensus con-
texts I and II Table 1 were not favourable in tobacco, leading to a significant decrease in LUC
activity of approximately 25 construct 12 or 50 construct 13.
We have also tested two gene contexts without A at − 3 and G at + 4 constructs 15 and 16.
Construct 15 corresponds to the AUG context found in a nodulin gene specifically expressed in
root nodules [33]. Construct 16 corresponds to another tissue-specific gene coding for a proline-
rich protein similar to the class of HyPRPs [34]. LUC activity obtained with these two constructs
was similar to that obtained with the anti-consen- sus construct 9, suggesting that these alternative
contexts are not appropriate, at least for the leaf mesophyll cells. These three contexts are also very
rare in vertebrate mRNA [10].
Table 1 Sequence, description and frequencies of the 17 AUG contexts tested
Frequency in dicot Frequency in monocot Description
Contruct AUG context
database
a
database
a
1 Consensus dicot I
3.914 AAAA AUG
GC 2
AGAA AUG Derives from consensus dicot I G at −3 0.783
GC AUAA AUG
0.065 3
Derives from consensus dicot I U at −3 GC
0.160 4
ACAA AUG Derives from consensus dicot I C at −3 0.718
GC 0.160
5 0.652
AAAC AUG Derives from consensus dicot I C at −1 corre- GC
sponds to consensus monocot III 6
2.609 2.083
Consensus dicot II C at −2 AACA AUG
GC AGCA AUG
2.244 7
Derives from consensus dicot II G at −3 1.826
GC 0.481
1.370 Derives from consensus dicot II U at −3
8 AUCA AUG
GC UCGU AUG
9 Anti-consensus dicot
CU 10
UCGU AUG Anti-upstream consensus dicot GC
AAAA AUG 11
Anti-downstream consensus dicot 0.065
CU AACC AUG
1.603 12
Consensus monocot I 0.913
GC AGCC AUG
2.083 13
Consensus monocot II 0.652
GC 0.481
0.130 Derives from consensus monocot C at −3
14 ACCC AUG
GC UCCU AUG Nodulin genbank L22765
0.130 15
CC 16
0.065 GUUG
Proline-rich protein genbank L20755 AUG AA
Negative control DAUG
17
a
Database described in Section 2.
Fig. 1. Relative LUC activity in extracts from tobacco leaf protoplasts after electroporation A, tobacco suspension cells
after biolistic transformation B, tobacco leaf cells after biolistic transformation C, maize suspension cells after bi-
olistic transformation D and from Norway spruce suspen- sion cells after biolistic transformation. LUC activity of each
construct was related to that of construct 1, considered as 100 for tobacco, construct 12 for maize and construct 5 for
Norway spruce, respectively. In C, only constructs 1 and 2 were tested.
cells are undifferentiated and actively divide. In this case, the constructs were introduced by biolis-
tic methods. The overall expression pattern Fig. 1B was
roughly similar to that for leaf except that the dicot consensus I construct 1 was not so conspic-
uous. Constructs 2 and 3 were as efficient. To check whether this difference was due to the trans-
formation system protoplast electroporation ver- sus intact cell bombardment or rather the plant
material leaf versus culture cells, we also trans- formed tobacco leaves with constructs 1 and 2,
using particle bombardment Fig. 1C. LUC activ- ity was significantly lower for construct 2 in both
leaf protoplast electroporation and intact leaf bombardment, while this was not the case for cell
culture bombardment. We could therefore at- tribute the variation in the relative expression
between constructs 1 and 2 to the plant material.
Another significant difference between the sus- pension and leaf cells was the relative greater
importance of the GC downstream consensus con- text in the former compare constructs 10 and 11
of Fig. 1A and B.
3
.
5
. Assay of the AUG context in maize suspension cells
Although initially designed to be analysed in a dicot species, the various constructs were also
introduced by particle bombardment into suspen- sion cells from maize, a monocot species Fig.
1D. The highest activity was obtained with con- structs 12 and 13, with the monocot consensus
sequence I and II, respectively. In contrast to tobacco leaf protoplasts, but as with tobacco sus-
pension cells, either A or G at position − 3 pro- duced high LUC activity. Another difference was
that the presence of C at positions − 1 and − 2 was necessary to support a high expression level
compare construct 12 and constructs 5 or 6. The lowest expression B 20 was obtained, as in
tobacco, with constructs 9, 15 and 16. This is not surprising, as the anti-consensus sequence is ap-
propriate to both plant classes. Both the anti-up- stream or downstream consensus affected LUC
activity, significantly reduced to approximately 30. No significant difference could be found
between the other constructs, with an average LUC activity close to 50 of that obtained with
constructs 12 or 13.
3
.
4
. Assay of the AUG context in tobacco suspension cells
The preference for a given initiation AUG con- text might be different for various tissues or at
distinct physiological states. To test this hypothe- sis, we expressed the same constructs in tobacco
cells from the BY2 suspension culture [26]. These
3
.
6
. Assay of AUG context in Norway spruce suspension cells
Finally, we bombarded the 17 constructs in embryogenic Norway spruce cell cultures to study
the importance of AUG context in a gymnosperm plant. When this study was initiated, only a small
number of gymnosperm gene sequences was avail- able in the databases, thus preventing a consensus
from being determined. We had therefore no ex- pectations for the relative expression level of the
different constructs as was the case for tobacco and maize. However, Joshi et al. [11] deduced a
consensus context from 93 non-angiosperm higher plants
bryophytes, pteridophytes
and gym-
nosperms: AG a ac ATG G C. Construct 5, which has a sequence corresponding to this con-
sensus, gave the highest level of LUC expression in Norway spruce cells Fig. 1E. While the anti-dicot
consensus construct 9 produced less than 25 relative LUC activity, the comparison of anti-up-
stream construct 10 and anti-downstream con- struct 11 consensus sequences showed that only
the upstream sequence was important.
4. Discussion