PII: S0014-5793(01)02246-3

E¤cient GFP mutations profoundly a¡ect mRNA transcription and

translation rates

Andrea Sacchetti, Tarek El Sewedy

, Ashraf F. Nasr, Saverio Alberti*

Laboratory of Experimental Oncology and Biotech Group, Department of Cell Biology and Oncology,

Instituto di Ricerche Farmacologiche Mario Negri, Consorzio Mario Negri Sud, 66030 Santa Maria Imbaro (Chieti), Italy

Received 5 January 2001; accepted 6 February 2001

First published online 23 February 2001

Edited by Julio Celis

Abstract Green fluorescent protein (GFP) variants with higher

expression efficiencies have been generated by mutagenesis.

Favorable mutations often improve the folding of GFP. However,

an effect on protein folding fails to explain the efficiency of

several other GFP mutations. In this work, we demonstrate that

mutations of the GFP open reading frame and untranslated

regions profoundly affect mRNA transcription and translation

efficiencies. The removal of the GFP 5PP untranslated region

halves the transcription rate of the GFP gene, but hugely

improves its translation rate. Mutations of the GFP open reading

frame or the addition of peptide sequences differentially reduce

the GFP mRNA transcription rate, translation efficiency and

protein stability. These previously unrecognized effects are

demonstrated to be critical to the efficiency of GFP mutants.

These findings indicate the feasibility of generating more

efficient GFP variants, with optimized mRNA transcription

and translation in eukaryotic cells. ß 2001 Federation of Euro-

pean Biochemical Societies. Published by Elsevier Science B.V.

Key words: Green £uorescent protein; RNA transcription

rate; RNA translation e¤ciency; Protein stability; Protein

folding; Gene mutation

1. Introduction

Green £uorescent protein (GFP) is a spontaneously £uores-

cent protein that is widely used as a recombinant protein tag

in living cells [1^3]. However, the expression of GFP is sub-

optimal in several experimental systems [4], often because of a

low folding e¤ciency at high temperatures [4,5]. Several GFP

mutants have thus been produced that demonstrate better

£uorescence and generate higher protein levels than the

wild-type (wt) [6^11]. Quite a few of these also show an im-

proved protein folding [6,8,10,11]. However, it was open to

question if all of the mutations with major e¡ects on GFP

expression levels a¡ected protein folding or protein stability,

or if they acted through other, as yet unrecognized, mecha-

nisms [4^6,8].

Little is known about the e¡ects of structural alterations of

the GFP gene on its transcription and translation abilities

[12]. In most genes, the latter are in£uenced by the 5P and

3P untranslated regions (UTs) of the mRNA [13^18], and by

sequences surrounding the translation start site, such as the

Kozak motif [19^23]. Sequence elements in the coding region

also a¡ect the e¤ciency of transcription, the mRNA stability

and the mRNA translation rate [24^29]. Thus, we investigated

if alterations in the GFP UT and open reading frame (ORF)

(deletions, point mutations and added peptide tags, as test

sequences for fusion to heterologous proteins [30,31]) could

signi¢cantly modify the GFP mRNA transcription and trans-

lation rates, or mRNA stability. Our ¢ndings demonstrate

that mutations of the GFP ORF and UT have major conse-

quences on mRNA transcription and translation. These pre-

viously unrecognized e¡ects are demonstrated to be critical to

the e¤ciency of GFP mutants. These results indicate the fea-

sibility of more e¤cient GFP expression constructs, and of

speci¢c mutagenesis and screening programs aimed at opti-

mizing mRNA transcription and translation in eukaryotic

cells.

2. Materials and methods

2.1. Vector construction

The full-length wtGFP cDNA (p10.1 plasmid) [12] was utilized to

generate the wtGFP expression constructs with or without UTs. The

wtGFP, and the S65T [6], Bex1 and Vex1 [7,32] mutants were ampli-

¢ed by PCR and subcloned in the pRK-5 expression vector [33] (Fig.

1). N-terminal (wtGFP-myc) and C-terminal (wtGFP-SB) tagged

wtGFP were also generated. All constructs were sequenced to ascer-

tain the absence of PCR-induced mutations.

2.2. Transfection of the GFP expression constructs

293T cells were cultured in DMEM with 10% fetal calf serum

(Gibco-BRL, Paisley, UK) and 200 Wg/ml geneticin (Sigma Chemical

Co., St. Louis, MO, USA). Cells were seeded in 10 cm diameter dishes

(Nunclon, Nunc, Denmark), and transfected by calcium phosphate

co-precipitation, as previously described [34].

2.3. Northern blot analysis

RNA for Northern hybridization was extracted from transfected

293T cells. The GFP mRNA stability was quanti¢ed as described

previously [33]. Brie£y, transfected cells were treated with 10 Wg/ml

actinomycin D to stop RNA polymerase II transcription. Total RNA

was extracted at di¡erent times after the treatment and analyzed by

Northern blotting. The levels of the GFP transcripts were measured

by densitometry, and analyzed with NIH Image 1.62, using a Kodak

gray scale for internal standardization (http://www.kodak.com/coun-

try/US/en/motion/postProduction/tools/taf.shtml). mRNA levels were

plotted against time (Fig. 2), and used to determine the mRNA half-

life (Table 1). The ratio between steady-state mRNA levels and RNA

stability was used to quantify the mRNA transcription rate. Since the

mRNA stability of the di¡erent constructs was found to be essentially

identical, the levels of mRNA were used as direct indicators of

mRNA transcription rates (Table 1).

PII: S0014-5793(01)02246-3

*Corresponding author. Fax: (39)-0872-570 412.

E-mail: [email protected]

Present address: King Abdulaziz City for Science and Technology,

Riyadh, Saudi Arabia.

FEBS 24684 5-3-01

FEBS 24684 FEBS Letters 492 (2001) 151^155

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by Elsevier - Publisher Connector

2.4. Western blot analysis

Pre-cleared cell lysates were run on 12% SDS^PAGE and trans-

ferred to nitrocellulose ¢lters. The ¢lters were treated with rabbit

anti-GFP (Clontech) and peroxidase-conjugated anti-rabbit antisera

(Calbiochem, La Jolla, CA, USA) essentially as described previously

[35]. Antibody-binding was revealed by chemiluminescence (ECL,

Amersham, Buckinghamshire, UK) (Fig. 3). The GFP protein stabil-

ity was quanti¢ed after blocking protein synthesis with cycloheximide

(25 Wg/ml). Cells were lysed at di¡erent times after the treatment, and

analyzed by Western blotting (Fig. 3). The GFP levels (P) at the

di¡erent time points were measured by densitometry, analyzed with

NIH Image 1.62, and used to determine the protein half-life (Table 1).

The mRNA translation rates of the di¡erent constructs (how many

molecules of GFP are synthesized per mRNA molecule) were

calculated as the ratio between corresponding protein and mRNA

levels. To correct for the di¡erent stabilities of the encoded pro-

teins, the protein levels were normalized versus the protein half-lives

(P/Pt

1=2

). As for the mRNA transcription rates, there was no need to

normalize for mRNA stability, since this was constant among the

di¡erent constructs (Fig. 2).

2.5. Fluorescence analysis

The GFP £uorescence analysis was performed by £ow cytometry

[34] (FACStar and Vantage, Becton-Dickinson, Synnyvale, CA, USA)

(Fig. 3), and by spectro£uorimetry (CM1T11I Spex spectro£uorime-

ter, Spex Industries, Edison, NJ, USA). Blue-excited GFPs are better

chromophores [36] than the wtGFP by 1.7-fold. However, the wtGFP

is excited about 3.5-fold less at 488 nm than at 396 nm [37,38]. Thus,

blue-excited GFPs emit around 6-fold more light than wtGFP when

excited at 488 nm in £ow cytometry. Fluorescence half-lives essentially

coincided with those of the corresponding peptides reported in Table

3. Results and discussion

The wtGFP, and ¢rst (S65T) [6] and second (Bex1 and

Vex1) [7] generation GFP mutants were expressed in mamma-

lian cells and analyzed. These constructs were designed to

include either or both the 5P and 3P UTs of the GFP gene,

N- or C-terminal GFP amino acidic tags, di¡erent Kozak

initiation of translation sequences [21], and di¡erent muta-

tions of the GFP ORF (Fig. 1 and Table 1). The S65T con-

struct bears this single mutation [6], whereas Bex1 and Vex1

bear multiple ORF mutations (S65T, V163A for Bex1;

V163A, S202F, T203Y for Vex1) [7,32]. All constructs were

transiently transfected into mammalian cells. The GFP

mRNA transcription and translation rates and stabilities

were measured, together with the GFP protein and £uores-

cence levels and stabilities (Figs. 2 and 3). The ratio between

steady-state mRNA levels and RNA stability was used to

quantify the GFP mRNA transcription rates. Since the

mRNA stability was found constant for the di¡erent GFP

variants, the mRNA levels were used as direct indicators of

the e¤ciency of mRNA transcription (Table 1). The ratio

between steady-state protein and mRNA levels was used to

quantify the mRNA translation rates. This ratio was normal-

ized versus the protein half-life, to correct for the di¡erential

e¡ects of each mutation on protein stability. Table 1 presents

the results obtained in transfected 293T cells. Analogous re-

sults were obtained in transfected COS-7 and 293 cells, sup-

porting the widespread validity of these ¢ndings.

3.1. UTs

The deletion of the 5P UT from wtGFP caused a 50% re-

duction in the GFP mRNA levels (Table 1). Since the stability

of the v5P UT mRNA was not a¡ected (Table 1 and Fig. 2),

this indicated that the removal of the 5P region of the gene

reduced the GFP mRNA transcription rate by 50%. On the

other hand, the deletion of the 5P UT led to a 200-fold in-

crease in the GFP protein levels (Table 1 and Fig. 3; wtGFP-

v5P UT versus wtGFP) [12]. The wtGFP-v5P UT and wtGFP

constructs expressed identical, native GFP proteins, and these

were con¢rmed to have an identical half-life (Table 1). Thus,

wtGFP-v5P UT mRNA possesses a 200-fold higher transla-

tion rate than wtGFP mRNA. The wtGFP-v5P UT and

wtGFP share identical Kozak initiation of translation sequen-

ces (Table 1 and see below) [21]. Thus, potent translation

regulatory elements other than the Kozak motif exist within

the 5P UT of GFP. These may impair ribosome scanning [39]

Fig. 1. Engineered GFP variants. wtGFP: full-length wtGFP cDNA in germline con¢guration. This includes the 5P and 3P UTs. wtGFP-v5P

UT: wtGFP that includes the 3P UT, but is devoid of the 5P UT; GFP-myc: N-terminal myc-tagged GFPs (wt, S65T, Vex1 or Bex1), devoid

of the 5P and 3P UTs; GFP-SB: C-terminal streptavidin-binding tagged wtGFP, devoid of the 5P and 3P UTs. UT: untranslated region; bent

arrow: ATG translation start codon; stop: stop codon; MYC : myc tag; SB: streptavidin-binding tag.

FEBS 24684 5-3-01

A. Sacchetti et al./FEBS Letters 492 (2001) 151^155152

or bind negative modulators of GFP translation [29,40], and

their removal allows the e¤cient translation of the GFP

mRNA. On the other hand, all of the mRNA devoid of a

3P UT displayed a lower translation rate than the wtGFP-

v5P UT, suggesting a translation-enhancing e¡ect of the 3P

UT of GFP [16,18].

3.2. Point mutations

Mutations in the GFP ORF were found to a¡ect the trans-

lation rate of the corresponding mRNA. The V163A mutation

was shown to cause a reduction in the translation rate of the

GFP mRNA by almost one-half (Table 1: compare Bex1-myc

with S65T-myc). This result was rather unexpected, since

Bex1-myc is an e¤cient V163A mutant [7]. However, a sim-

ilarly low translation rate was demonstrated by Vex1-myc

(Table 1). The ¢nding that a single codon change in the

GFP ORF is su¤cient to profoundly a¡ect the GFP trans-

lation rate is novel. It would be of interest to determine if this

single codon mutation directly a¡ects translation, e.g. by

translation pausing due to RNA secondary structures [28],

or if it induces the binding of translation-inhibitory factors

[29].

3.3. GFP tags

GFP is routinely used as a fusion partner to heterologous

proteins, and these have been demonstrated to heavily a¡ect

GFP function [5,30,41]. GFP tags known to a¡ect the folding

of GFP [41] were used in this work as examples of heterolo-

gous fusion partners. Remarkably, the SB tag was found to

a¡ect both mRNA transcription and translation rates, con-

¢rming the role of coding sequences in regulating the mRNA

transcription e¤ciencies [24] and translation rates [28,29].

The wtGFP-SB and wtGFP-myc bear Kozak initiation of

translation sequences of di¡erent strength [21] (Table 1).

However, di¡erent Kozak sequences are unlikely to account

for major di¡erences in GFP translation e¤ciencies. Indeed,

wtGFP-v5P UT demonstrates the highest translation rate

among the constructs tested in this work. In particular, it is

twice as e¤cient as wtGFP-myc, in spite of the largely sub-

optimal Kozak sequence of the former versus the `perfect'

Kozak motif in the latter [21,42]. The 200-fold better trans-

lation of wtGFP-v5P UT compared to that of the wtGFP, in

spite of an identical germline Kozak motif, is also consistent

with an overall minor role in GFP translation (Table 1, Fig. 3

and see above). Kozak motifs were ¢rst identi¢ed by compar-

ison of numerous eukaryotic mRNA sequences that surround

the AUG start codon [43]. However, the relative strength of

di¡erent Kozak sequences was experimentally assessed largely

in expression systems for preproinsulin or chloramphenicol

acetyl transferase [21]. Thus the strength of Kozak sequences

may vary depending on their sequence context [44]. In other

words, mRNA elements other than the Kozak motif contrib-

ute to the determination of their overall translation compe-

tence, and the relative importance of these di¡erent elements

probably varies among di¡erent mRNAs.

wtGFP-SB, Bex1-myc and Vex1-myc demonstrate the high-

est folding levels (up to 80% of the GFP molecules) [41] and

the lowest translation rates among the GFP variants analyzed

(Table 1). This raised the possibility that high folding levels

were mechanistically linked to a `slow' translation [45,46].

However, a comparison of wtGFP and wtGFP-v5P UT, which

are identical wtGFP molecules with widely di¡erent mRNA

translation rates, demonstrates essentially identical protein

folding abilities (25% of the GFP molecules [41]). Thus, the

GFP folding e¤ciency appears to be a fundamental property

of speci¢c peptide sequences, and is unlikely to be causally

related to translational `speed'.

Fig. 2. GFP mRNA stability in transfected 293T cells. mRNA syn-

thesis was blocked with actinomycin D and the relative residual

amounts of GFP mRNA were quanti¢ed at di¡erent time points.

wtGFP: squares ; wtGFP-v5P UT: triangles; wtGFP-myc: circles.

Table 1

GFP mRNA stabilities and transcription/translation rates

GFP form mRNA levels

mRNA t

1=2

Translation rate

Kozak sequence

wtGFP 100 7 1 54 1 caaagATGa

wtGFP-v5P UT 50 7 170 54 210 caaagATGa

wtGFP-SB 80 30 33 38 cgcatATGa

wtGFP-myc 100 7 100 33 100 agaccATGg

S65T-myc 100 80 25 103 agaccATGg

Bex1-myc 100 60 32 62 agaccATGg

Vex1-myc 100 40 25 55 agaccATGg

The results are average of three independent experiments. All values, except for half-lives, are normalized versus wtGFP-myc and expressed as

percentage.

GFP mRNA levels.

GFP mRNA half-life expressed in hours.

GFP protein levels (P).

GFP protein half-life expressed in hours.

mRNA translation rates.

Kozak ribosomal binding sequence [21] ; the ATG start of translation is in capital letters.

FEBS 24684 5-3-01

A. Sacchetti et al./FEBS Letters 492 (2001) 151^155 153

4. Conclusions

In the past, structural and functional analyses of GFP mu-

tants identi¢ed in screening programs have repeatedly raised

questions as to their actual mechanisms of action [4,6,8]. Our

results demonstrate that several commonly engineered altera-

tions in GFP structure have remarkable consequences on

mRNA expression parameters. The major a¡ected mechanism

is the e¤ciency of mRNA translation, that was shown to span

a 200-fold range. Signi¢cant e¡ects were also demonstrated on

protein stability and mRNA transcription rates. These previ-

ously unrecognized e¡ects were found to be critical to the

overall e¤ciency of GFP mutants (Table 1; compare the

translation rates of the di¡erent constructs and the corre-

sponding GFP protein levels). Thus, a strategy of optimiza-

tion of GFP mRNA expression parameters may contribute

considerably to the improvement of GFP expression. The en-

gineering of the 5P and 3P UTs of GFP, in particular, might

permit the reaching of both higher transcription and trans-

lation rates. If GFP mutations are found to have similar con-

sequences on GFP mRNA transcription and translation rates

in lower eukaryotes, high throughput screening strategies for

more e¤cient GFP mutants could be e¤ciently designed in

commonly used yeast strains [47,48].

Acknowledgements: This work was supported by the Italian Associa-

tion for Cancer Research (AIRC), the Italian National Research

Council (Convenzione CNR, Consorzio Mario Negri Sud), and the

Consorzio per la Medicina Tropicale (CMT). A.S. is recipient of a

fellowship from Italian Foundation for Cancer Research (FIRC).

T.E.S. and A.F.N. were recipients of fellowships from the CMT.

References

[1] Prasher, D.C. (1995) Trends Genet. 11, 320^323.

[2] Cubitt, A.B., Heim, R., Adams, S.R., Boyd, A.E., Gross, L.A.

and Tsien, R.Y. (1995) Trends Biochem. Sci. 20, 448^455.

[3] Ludin, B. and Matus, A. (1998) Trends Cell Biol. 8, 72^77.

[4] Tsien, R.Y. (1998) Annu. Rev. Biochem. 67, 509^544.

[5] Sacchetti, A. and Alberti, S. (1999) Nat. Biotechnol. 17, 1046.

[6] Heim, R., Cubitt, A.B. and Tsien, R.Y. (1995) Nature 373, 663^

664.

[7] Anderson, M.T., Tjioe, I.M., Lorincz, M.C., Parks, D.R., Her-

zenberg, L.A. and Nolan, G.P. (1996) Proc. Natl. Acad. Sci.

USA 93, 8508^8511.

[8] Heim, R. and Tsien, R.Y. (1996) Curr. Biol. 6, 178^182.

[9] Crameri, A., Whitehorn, E.A., Tate, E. and Stemmer, W.P.C.

(1996) Nat. Biotechnol. 14, 315^319.

[10] Cormack, B.P., Valdivia, R.H. and Falkow, S. (1996) Gene 173,

33^38.

[11] Siemering, K.R., Golbik, R., Sever, R. and Haselo¡, J. (1996)

Curr. Biol. 6, 1653^1663.

[12] Chal¢e, M., Tu, Y., Euskirchen, G., Ward, W.W. and Prasher,

D.C. (1994) Science 263, 802^805.

Fig. 3. Expression levels of the transfected GFP constructs. 293T cells were transfected with (A) wtGFP; (B) wtGFP-myc; (C) Bex1-myc; (D)

wtGFP-v5P UT; (E) wtGFP-SB; (F) S65T-myc. Panels show log £uorescence pro¢les of 5000 cells. Solid line: transfected cells. Dotted line:

vector-alone transfected cells. Insets : GFP protein levels, as determined by Western blot analysis. The GFP protein levels in (A) were too low

to be shown on scale with the other samples.

FEBS 24684 5-3-01

A. Sacchetti et al./FEBS Letters 492 (2001) 151^155154

[13] Pantopoulos, K., Johansson, H.E. and Hentze, M.W. (1994)

Prog. Nucleic Acids Res. Mol. Biol. 48, 181^238.

[14] Tanguay, R.L. and Gallie, D.R. (1996) Mol. Cell. Biol. 16, 146^

156.

[15] Al-Qahtani, A. and Mensa-Wilmot, K. (1996) Nucleic Acids Res.

24, 1173^1174.

[16] Munroe, D. and Jacobson, A. (1990) Mol. Cell. Biol. 10, 3441^

3455.

[17] Hann, L.E., Webb, A.C., Cai, J.-M. and Gehrke, L. (1997) Mol.

Cell. Biol. 17, 2005^2013.

[18] Bailey-Serres, J. and Dawe, R.K. (1996) Plant Physiol. 112, 685^

695.

[19] Iida, Y. and Masuda, T. (1996) Nucleic Acids Res. 24, 3313^

3316.

[20] Le, S.Y. and Maizel, J.V.J. (1997) Nucleic Acids Res. 25, 362^

369.

[21] Kozak, M. (1991) J. Biol. Chem. 266, 19867^19870.

[22] Sachs, A.B., Sarnow, P. and Hentze, M.W. (1997) Cell 89, 831^

838.

[23] Afshar-Kharghan, V., Li, C.Q., Khoshnevis-Asl, M. and Lopez,

J.A. (1999) Blood 94, 186^191.

[24] Xiang, S., Parsons, H.K. and Murray, M. (1998) Gene 209, 123^

129.

[25] Bernstein, P.L., Herrick, D.J., Prokipcak, R.D. and Ross, J.

(1992) Genes Dev. 6, 642^654.

[26] Veyrune, J.L., Carillo, S., Vie, A. and Blanchard, J.M. (1995)

Oncogene 11, 2127^2134.

[27] Ross, J. (1995) Microbiol. Rev. 59, 423^450.

[28] Zama, M. (1999) Nucleic Acids Symp. Ser. 42, 81^82.

[29] Xu, Y.H. and Grabowski, G.A. (1999) Mol. Genet. Metab. 68,

441^454.

[30] Waldo, G.S., Standish, B.M., Berendzen, J. and Terwilliger, T.C.

(1999) Nat. Biotechnol. 17, 691^695.

[31] Keiler, K.C., Waller, P.R. and Sauer, R.T. (1996) Science 271,

990^993.

[32] Sacchetti, A., Ciccocioppo, R. and Alberti, S. (2000) Histol. His-

topathol. 15, 101^107.

[33] Sambrook, J., Fritsch, E.F. and Maniatis, T. (1989) Cold Spring

Harbor Laboratory, New York.

[34] Dell'Arciprete, R., Stella, M., Fornaro, M., Ciccocioppo, R.,

Capri, M.G., Naglieri, A.M. and Alberti, S. (1996) J. Histochem.

Cytochem. 44, 629^640.

[35] El-Sewedy, T., Fornaro, M. and Alberti, S. (1998) Int. J. Cancer

75, 324^331.

[36] Matz, M.V., Fradkov, A.F., Labas, Y.A., Savitsky, A.P., Za-

raisky, A.G., Markelov, M.L. and Lukyanov, S.A. (1999) Nat.

Biotechnol. 17, 969^973.

[37] Patterson, G.H., Knobel, S.M., Sharif, W.D., Kain, S.R. and

Piston, D.W. (1997) Biophys. J. 73, 2782^2790.

[38] Ward, W.W., Prentice, H.J., Roth, A.F., Cody, C.W. and

Reeves, S.C. (1982) Photochem. Photobiol. 35, 803^808.

[39] Beyer, D., Skripkin, E., Wadzack, J. and Nierhaus, K.H. (1994)

J. Biol. Chem. 269, 30713^30717.

[40] Dubnau, J. and Struhl, G. (1996) Nature 379, 694^699.

[41] Sacchetti, A., Cappetti, V., Marra, P., Dell'Arciprete, R., El Sew-

edyl, T., Crescenzi, C. and Alberti, S. (2000) J. Cell. Biochem.

[42] Kozak, M. (1986) Cell 44, 283^292.

[43] Kozak, M. (1984) Nucleic Acids Res. 12, 857^872.

[44] Kozak, M. (1989) Mol. Cell. Biol. 9, 5073^5080.

[45] Fedorov, A.N. and Baldwin, T.O. (1997) J. Biol. Chem. 272,

32715^32718.

[46] Harding, H.P., Zhang, Y. and Ron, D. (1999) Nature 397, 271^

274.

[47] Cereghino, G.P. and Cregg, J.M. (1999) Curr. Opin. Biotechnol.

10, 422^427.

[48] Sudbery, P.E. (1996) Curr. Opin. Biotechnol. 7, 517^524.

FEBS 24684 5-3-01

A. Sacchetti et al./FEBS Letters 492 (2001) 151^155 155