JCat: a novel tool to adapt codon usage of a target
gene to its potential expression host
Andreas Grote1,2
, Karsten Hiller1
, Maurice Scheer1,3
, Richard Mu¨nch1
, Bernd No¨rtemann2
,
Dietmar C. Hempel2
and Dieter Jahn1,
*
1
Institut fu¨r Mikrobiologie, Spielmannstraße 7, 2
Institut fu¨r Bioverfahrenstechnik, Gaußstraße 17, Technische
Universita¨t Braunschweig, D-38106 Braunschweig, Germany and 3
Fachbereich fu¨r Informatik, Am Exer 2,
Fachhochschule Wolfenbu¨ttel, D-38302 Wolfenbu¨ttel, Germany
Received February 11, 2005; Revised and Accepted March 7, 2005
ABSTRACT
A novel method for the adaptation of target gene
codon usage to most sequenced prokaryotes and
selected eukaryotic gene expression hosts was
developed to improve heterologous protein production.
In contrast to existing tools, JCat (Java Codon
Adaptation Tool) does not require the manual definition
of highly expressed genes and is, therefore, a
very rapid and easy method. Further options of
JCat for codon adaptation include the avoidance of
unwanted cleavage sites for restriction enzymes
and Rho-independent transcription terminators. The
output of JCat is both graphically and as Codon
Adaptation Index (CAI) values given for the pasted
sequence and the newly adapted sequence.
Additionally, a list of genes in FASTA-format can be
uploaded to calculate CAI values. In one example,
all genes of the genome of Caenorhabditis elegans
were adapted to Escherichia coli codon usage and
further optimized to avoid commonly used restriction
sites. In a second example, the Pseudomonas aeruginosa
exbD gene codon usage was adapted to E.coli
codon usage with parallel avoidance of the same
restriction sites. For both, the degree of introduced
changes was documented and evaluated. JCat is
integrated into the PRODORIC database that hosts
all required information on the various organisms
to fulfill the requested calculations. JCat is freely
accessible at http://www.prodoric.de/JCat.
INTRODUCTION
The genetic code is degenerated, e.g. there are multiple
codons, up to six different ones, coding for the same amino
acid. Depending on the employed codon usage of each organism,
various codons are employed differentially (1,2). For
example, GC-rich organisms prefer GC-containing codons
over AT-containing ones. Therefore, according to the nonrandomness
of the codon usage, there are optimal and nonoptimal
codons for each organism. This so-called ‘dialect’ of
codon choice is not universal and unique for each organism.
The codon usage of different genes of one organism also
relates to their specific rate of expression (3). Currently,
one of the central techniques used in biomedical research
and biotechnological production processes is the heterologous
gene expression. In order to succeed in the expression of genes
outside their natural context, several problems were experienced
(4). Codon usage, one major factor among others, has
a significant impact on heterologous gene expression. Rarely
employed codons of the expression host found in the target
gene can lead to poorly translated mRNAs, decreased mRNA
stability and sometimes to premature termination of translation
(5,6). Finally, the usage of rare codons for arginine in
Escherichia coli can provoke misincorporation of amino acids
(7). Currently, there are two ways to prevent the problems
described above. The first method is to supplement the
gene expression host with the tRNAs that are otherwise infrequent
in this organism (8). The second method is to adapt the
codon usage of the gene that should be expressed in the host
organism (9). Sharp and Li (10) developed the Codon Adaptation
Index (CAI). This index permits the calculation of a
comparable value for codon usage. The calculation of the CAI
requires the definition of highly expressed genes. The codon
usage of other genes from this organism is calculated with
respect to this subset of genes. The CAI is the prevailing
empirical measure of expressivity.
To our knowledge, there are currently only two bioinformatic
tools available for the adaptation of codon usage. The first
tool is UpGene (11). This tool is truly valuable for the codon
optimization of SIV/HIV coding sequences in order to maximize
their expression in eukaryotic cells. However, with this
*To whom correspondence should be addressed. Tel: +49 531 391 5801; Fax: +49 531 391 5854; Email: d.jahn@tu-bs.de
ª The Author 2005. Published by Oxford University Press. All rights reserved.
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access
version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press
are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but
only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact journals.permissions@oupjournals.org
W526–W531 Nucleic Acids Research, 2005, Vol. 33, Web Server issue
doi:10.1093/nar/gki376
tool it is not possible to adapt codon usage to other organisms.
Another disadvantage is the fact that UpGene is only available
as a stand-alone application that makes its immediate use
complicated. The second tool is Codon Optimizer (12). This
tool is again only available as a stand-alone application. Since
this tool uses CAI as the basic value for codon adaptation, it
requires a set of highly expressed genes to be defined manually
in order to get an optimal codon optimization. Here, simplicity
stands against universality.
Here, we describe a novel and easy-to-use program, called
JCat (Java Codon Adaptation Tool), for the adaptation of
codon usage to most prokaryotic and some eukaryotic organisms
of biotechnological interest. JCat is presented as a web
service that offers immediate program usage. The web interface
is easy to understand and the calculations are performed
in real time. Furthermore, the algorithm does not require a
definition of highly expressed genes. Calculations are made in
advance with the aid of an algorithm proposed by Carbone
et al. (13). The results of the calculation are stored in the
PRODORIC database, where data of most freely available
genomes of sequenced prokaryotes are hosted (14). The
PRODORIC database is frequently updated; therefore, JCat
is always up to date. Additionally, JCat features the possibility
of avoiding cleavage sites for certain restriction enzymes
and includes the avoidance of Rho-independent transcription
terminators in the codon-optimized DNA sequence.
The algorithm for the prediction of Rho-independent
transcription terminators was previously proposed by
Ermolaeva et al. (15).
MATERIALS AND METHODS
Algorithm for the codon optimization
The algorithm is based on the calculation of the CAI (10).
Each codon is given a weight with respect to the subset of
highly expressed genes defined for the considered organism.
The so-called relative adaptiveness of a codon is defined as:
wi ¼
fi
fmax ið Þ
1
where fi is the frequency of a codon (i) and fmax(i) is the
frequency of the codon most often used to code for the considered
amino acid in the subset of highly expressed genes.
The CAI for a gene ‘g’ can be calculated according to
Equation 2:
CAIg ¼
YN
i¼1
wi
!1=N
2
where N is the number of codons in a gene ‘g’ without the
initiation and stop codons.
The calculation of the relative adaptiveness for all genomes
in the PRODORIC database was made in advance. The subset
Figure 1. Screen shot of the web interface for the codon adaptation tool JCat.
Nucleic Acids Research, 2005, Vol. 33, Web Server issue W527
of highly expressed genes for each organism was defined by
applying the algorithm proposed by Carbone et al. (13). The
algorithm is based on the assumption that in each genome
there is a set of genes with high codon bias. The algorithm
is iterative and reduces the set of genes (initially all genes of an
organism) during each iteration until only 1% of genes remain
with the highest codon bias of the initial set of genes.
The optimization of a given sequence splits into two parts.
First, the sequence is examined whether it is either a correct
gene sequence or a correct amino acid sequence. Subsequently,
depending on the type of sequence, it is translated
into an amino acid sequence. The second step is to translate the
amino acid sequence into a gene sequence by using the codons
that got the highest relative adaptiveness for the amino acid
in question. In this way, every amino acid of the sequence is
replaced until the whole sequence is retranslated.
Algorithm for the prevention of Rho-independent
transcription terminators
The algorithm for the prevention of Rho-independent transcription
terminators consists of two parts. The first part recognizes
Rho-independent transcription terminators. The second
part is dedicated to the improvement of the determined
sequence in order to avoid these structures.
The algorithm proposed by Ermolaeva et al. (15) is based
on the assumption that Rho-independent transcription terminators
have two major characteristics. The first one is a
defined mRNA stem–loop (hairpin) that can be predicted
with the help of an energy scoring function. The second
point is a poly-U strand close to the hairpin to allow translation
termination (16).
After recognition of stem–loop-like structures and poly-U
stretches, codons that are involved with both parameters
are adapted separately by replacing optimal codons in the
sequence by codons that are not optimal. Once transcription
terminators are no longer being found in the adapted sequence,
the sequence is stored. Each codon that is in the area of the
loop and the corresponding poly-U transcription termination is
exchanged. The decision as to which sequence has to be displayed
is made with respect to the CAI that is calculated for
each adapted sequence without transcription terminators.
Algorithm to prevent cleavage sites of specified
restriction endonucleases
The algorithm for the prevention of certain cleavage sites
resembles the algorithm described above. For restriction
enzymes, cleavage sites are derived from the REBASE database
(17). In a first step, cleavage sites are predicted. In a
Figure 2. Distribution of CAI values of the C.elegans genome under different conditions. (A) Calculation of the CAI applying the relative adaptiveness of C.elegans.
(B) Calculation of the CAI applyingthe relative adaptiveness of E.coli. (C) Distributionof CAI valuesof the genes from C.elegans after adaptationto the codonusage
of E.coli. The CAI was calculated applying the relative adaptiveness of E.coli. (D) Same as shown in (C); however, restriction endonucleases cleavage sites for ApaI,
AccI, BamHI, BstXI, ClaI, DraII, EcoO109I, EcoRV, EcoRI, EagI, HincII, HindIII, KpnI, NotI, PstI, SalI, SmaI, SpeI, SacII, SacI, XhoI and XbaI were eliminated,
resulting in a decrease of the CAI.
W528 Nucleic Acids Research, 2005, Vol. 33, Web Server issue
second step, the detected restriction site is eliminated. The
cleavage sites are removed using the same method as
described above. The optimal codons in the area of the cleavage
site are substituted by non-optimal codons. Then, the
considered cleavage site is searched again in the DNA.
Once no cleavage site is found in the sequence it is stored.
After the substitution of all codons in the area of the cleavage
site, the CAI, calculated for the various improved sequences,
decides which adapted sequence is displayed.
Web interface and analyses
The algorithms were programmed in Java (http://java.sun.com)
to take advantage of its object-oriented technology and to allow
integration into dynamic web sites using Java Server Page
(JSP) technology. Jakarta Tomcat runs as the servlet container
and web server on a Linux machine. Tomcat is available as an
open source tool at http://jakarta.apache.org/tomcat.
Graphical output of the data is carried out with the help of
the Java package JFreeChart. JFreeChart is freely available at
http://jfree.org/jfreechart.
Alignments were carried out with ClustalX that is freely
available at http://www.biolinux.org/clustalx.html.
RESULTS
Rational of the approach
In basic biochemical research and for various biotechnological
applications, heterologous protein production is of central
importance. However, there is only a limited number of prokaryotic
and eukaryotic production hosts for the wealth of
organisms under investigation. The codon usage of the gene
of interest and of its desired production host often differ significantly.
Frequently, this results in low protein recoveries. In
the age of commercial complete gene synthesis, a synthetic
target gene with perfect host codon usage offers an attractive
alternative to rare tRNA-supplemented strains, which are
currently only available for E.coli.
In this context, JCat offers the possibility to adapt the
codon usage of a gene of interest to the one of a selected
organism. For this purpose, the CAI for a list of genes or for
a pasted sequence is calculated and displayed. These values
provide the basis for the codon optimization. They are also
consulted during additional optimization steps, including
unwanted restriction site and transcription terminator prevention.
There is no need for the manual definition of codon
usage of highly expressed genes for the organism of interest.
Figure 3. Graphical presentation of the relative adaptiveness (wij) of the codon usage of the gene for the transport protein ExbD from P.aeruginosa to the codon usage
of E.coli. The mean codon usage of E.coli is presented as a blue line. The gray lines above and below the mean codon usage mark the standard deviation of the codon
usage. (A) Unadapted codon usage. (B) E.coli adapted codon usage including the elimination of various restriction enzyme cleavage sites.
Nucleic Acids Research, 2005, Vol. 33, Web Server issue W529
These have been precalculated for all published prokaryotic
and various eukaryotic organisms and are an integral part
of JCat.
USE OF THE JCAT WEB INTERFACE
Input
There are several options offered by JCat for codon optimization
of DNA sequences (Figure 1). First, the sequence is pasted
into a text field. In a second step, the pasted sequence is
specified as a DNA, an RNA or an amino acid sequence.
Further options for the codon adaptation are offered. Options
include the possibility to avoid Rho-independent transcription
terminators and certain cleavage sites of defined restriction
enzymes in the adapted sequence. Finally, the organism to
which the codon usage should be adapted is selected.
Alternatively, for the calculation of the CAI for a list of
genes, the web interface provides a menu item called ‘CAICalculation’.
The selection of this item opens a new webpage
with a new form. This form includes a field for file upload. The
file must be in valid FASTA-format. Sequences with other
letters than A, G, C, T and U are ignored. After selection
of the target organism for the calculation, all necessary settings
are made.
Output
The detailed nature of the output depends on the input options.
If a DNA/RNA sequence is to be analyzed, the CAI values are
displayed for the pasted sequence and for the optimized
sequence. Furthermore, a graphical representation of the relative
adaptiveness for each codon of the sequence is available
(Figure 3). Again, data for the analyzed and the adapted
sequence are shown.
If an amino acid sequence is analyzed, the results are only
presented for the adapted sequence.
The output of the calculated CAIs for a list of genes is in the
form of a two-column table containing the identifier of the
FASTA sequence and individual calculated CAIs.
In order to illustrate the mode of operation of JCat, the CAI
values for all genes of the whole Caenorhabditis elegans genome
were calculated (Figure 2A). Subsequently, the CAI for
the C.elegans genome was calculated with the relative adaptiveness
of E.coli to simulate the case of unadapted codon
usage (Figure 2B). One of the main differences between natural
codon usage of the genes of a certain organism (Figure 2A)
and unadapted codon usage of this organism (Figure 2B) is the
degree of diversity. The natural codon usage of the C.elegans
genome varied between 0.2 and 0.6. The unadapted codon
usage of the C.elegans genome in E.coli showed variation
Figure 4. Nucleotide sequence comparison of P.aeruginosa exbD before (pasted) and after (optimal) adaptation to E.coli codon usage, and with the elimination of
restriction sites (improved).
W530 Nucleic Acids Research, 2005, Vol. 33, Web Server issue
of the CAIs only between 0.1 and 0.3, which is besides that
of course lower.
Finally, the codons of the C.elegans genome were adapted
to the genome of E.coli. If the optimization is executed
without further options, the CAIs of all genes should be
shifted to a value of 1.0 (Figure 2C). In our example, we
further chose to optimize the analyzed DNA sequences of the
C.elegans genome for the avoidance of the cleavage sites of
different standard restriction enzymes (ApaI, AccI, BamHI,
BstXI, ClaI, DraII, EcoO109I, EcoRV, EcoRI, EagI, HincII,
HindIII, KpnI, NotI, PstI, SalI, SmaI, SpeI, SacII, SacI, XhoI
and XbaI). The restriction enzymes whose cleavage sites were
avoided are the restriction enzymes that are located on the
multiple cloning site of the widely spread vector pBluescript.
Figure 2D shows the efficiency of the algorithm. The codon
usage was optimized for each gene of the C.elegans genome to
E.coli codon usage with the exclusion of certain restriction
sites, demonstrating the efficient optimization of codon usage.
Most CAI values lay between 0.9 and 1.0.
However, the usual laboratory application of the JCat tool
might rather be codon optimization of single genes of interest.
Therefore, in a second example we adapted the codon usage of
the ExbD transport protein from Pseudomonas aeruginosa to
the codon usage of E.coli (Figure 3). The unadapted codon
usage of the P.aeruginosa exbD gene in E.coli is given in
Figure 3A. The codon usage adaptation again avoided the
restriction enzyme cleavage sites of the multiple cloning
site of pBluescript. In a first step, every codon was substituted
by the corresponding optimal codon of E.coli. In a second step,
the algorithm searched for cleavage sites. In this example,
cleavage sites were found in position 60 (HincII), 216
(HincII), 300 (PstI) and 318 (HincII) of the DNA. In a
third step, the codons in the area of these sites were substituted
by non-optimal codons. Figure 3B gives the graphical presentation
of the codon optimization. Obviously, the codons
with the number 22, 74, 102 and 108 have been exchanged by
non-optimal codons. These codons were in the area of the
cleavage sites. Figure 4 compares the original P.aeruginosa
DNA sequence with that optimized for E.coli demonstrating
the degree of optimization.
In addition to adaptation to E.coli codon usage demonstrated
here, every known gene can be adapted to the codon
usage of all prokaryotic organisms with completed genome, as
well as to those of selected eukaryotes, including Saccharomyces
cerevisae, Homo sapiens, Mus musculus, Drosophila
melanogaster and C.elegans. Finally, the codon usage of
homologous genes of a gene expression host might be subject
to optimization since optimal codon usage is not used throughout
the complete genome.
ACKNOWLEDGEMENTS
The authors thank Dr Barbara Schulz for critical proofreading
of the manuscript. This work was funded by the German
Bundesministerium fu¨r Bildung und Forschung (BMBF) for
the Bioinformatics Competence Center Imtergenomics and
Sonderforschungsbereich 578 ‘From gene to product’,
Deutsche Forschungsgemeinschaft; Fonds der Chemischen
Industrie. Funding to pay the Open Access publication charges
for this article was provided by BMBF.
Conflict of interest statement. None declared.
REFERENCES
1. Ikemura,T. (1981) Correlation between the abundance of Escherichia
coli transfer RNAs and the occurrence of the respective codons
in its protein genes. J. Mol. Biol., 146, 1–21.
2. Ikemura,T. (1985) Codon usage and tRNA content on unicellular and
multicellular organisms. Mol. Biol. Evol., 2, 13–34.
3. Gouy,M. and Gaultier,C. (1982) Codon usage in bacteria: correlation
with gene expressivity. Nucleic Acids Res., 10, 7055–7074.
4. Gustafsson,C., Govindarajan,S. and Minshull,J. (2004) Codon bias
and heterologous protein expression. Trends Biotechnol., 22,
346–353.
5. Sorensen,M.A., Kurland,C.G. and Pedersen,S. (1989) Codon usage
determine translation rate in Escherichia coli. J. Mol. Biol., 207,
365–377.
6. Goldman,E., Rosenberg,A.H., Zubay,G. and Studier,F.W. (1995)
Consecutive low-usage leucine codons block translation only when near
the 50
end of a message in Escherichia coli. J. Mol. Biol., 245,
467–473.
7. Calderone,T.L., Stevens,R.D. and Oas,T.G. (1996) High-level
misincorporation of lysine for arginine at AGA codon
in a fusion protein expressed in Escherichia coli. J. Mol. Biol., 262,
407–412.
8. Brinkmann,U., Mattes,R.E. and Buckel,P. (1989) High-level expression
of recombinant genes in Escherichia coli is dependent on the availability
of the dnaY gene product. Gene, 85, 109–114.
9. Hernan,R.A., Hui,H.L., Andracki,M.E., Noble,R.W., Sligar,S.G.,
Walder,J.A. and Walder,R.Y. (1992) Human hemoglobin expression in
Escherichia coli: importance of optimal codon usage. Biochemistry,
31, 8619–8628.
10. Sharp,P.M. and Li,W.H. (1987) The codon adaptation index—a measure
of directional synonymous codon usage bias, and its potential
applications. Nucleic Acid Res., 15, 1281–1295.
11. Gao,W., Rzewski,A., Sun,H., Robbins,P.D. and Gambotto,A. (2004)
UpGene: Application of a web-based DNA codon optimization
algorithm. Biotechnol. Prog., 20, 443–448.
12. Fuglsang,A. (2003) Codon optimizer: a freeware tool for codon
optimization. Protein Expr. Purif., 31, 247–249.
13. Carbone,A.,Zinovyev,A.andKepes,F.(2003)Codonadaptationindexas
a measure of dominating codon bias. Bioinformatics, 19,
2005–2015.
14. Munch,R., Hiller,K., Barg,H., Heldt,D., Linz,S., Wingender,E. and
Jahn,D. (2003) PRODORIC: prokaryotic database of gene regulation.
Nucleic Acids Res., 31, 266–269.
15. Ermolaeva,M.D., Khalak,H.G., White,O., Smith,H.O. and Salzberg,S.L.
(2000) Prediction of transcription terminators in bacterial genomes.
J. Mol. Biol., 301, 27–33.
16. d’Aubenton Carafa,Y., Brody,E. and Thermes,C. (1990) Prediction of
rho-independent Escherichia coli transcription terminators. A statistical
analysis of their RNA stem–loop structures. J. Mol. Biol., 216,
835–858.
17. Roberts,R.J., Vincze,T., Posfai,J. and Macelis,D. (2003) REBASE:
restriction enzymes and methyltransferases. Nucleic Acids Res.,
31, 418–420.
Nucleic Acids Research, 2005, Vol. 33, Web Server issue W531