Gene, 108 (1991) 167-174 0 1991 Elsevier Science Publishers

GENE

B.V. All rights reserved.

167

0378-l 119/91/$03.50

06148

Efficient gene expression in mammalian retroviral vector

cells from a dicistronic transcriptional unit in an improved

(Recombinant DNA; bicistronic; luciferase; neomycin phosphotransferase;

retrovirus; translation)

Fred Levine, Jiing Kuan Yee and Theodore Friedmann Department of Pediatrics, Center for Molecular Genetics, UCSD School of Medicine, La Jol[a, CA 92093-0634 (U.S.A.) Received by J.A.Hoch: 13 June 1991 Revised/Accepted: 30 July 1991/2 August Received at publishers: 23 August 1991

1991

SUMMARY

We have studied the properties of dicistronic transcriptional units in retroviral vectors. In these vectors, the promoter in the 5’ retroviral long terminal repeat (LTR) controls expression of both an upstream cistron (/UC)encoding firefly luciferase and a downstream cistron (neo), a selectable marker encoding neomycin phosphotransferase (NPTII). By assaying for simultaneous expression of luc and neo after transfection or infection of hamster BHK, rat 208F, and mouse retroviral packaging cell lines, we have identified important factors that tiect expression from the downstream cistron, including the presence of intercistronic ATG sequences, the length of the intercistronic sequence and conformity of the sequence surrounding the downstream start codon to the eukaryotic consensus sequence. Optimized dicistronic vectors produced amounts of NPTII comparable to a vector in which neo was driven by a strong internal promoter consisting of a modified Rous sarcoma virus LTR. Additionally, they produced higher virus titers and demonstrated improved stability of gene expression in the absence of selection. By virtue of their physical compactness and elimination of the need for a separate promoter for every gene, dicistronic transcriptional units allow the introduction of larger genes into retroviral vectors and may allow for more than two genes to be placed in a single vector.

INTRODUCTION

The vast majority of eukaryotic genes contain a single translational unit under the control of a single promoter (Kozak, 1989). Exceptions are provided primarily by some viral genes such as the retroviralgag and@ genes in which

Correspondence to: Dr. T. Friedmann, Center for Molecular Genetics, UCSD School of Medicine, La Jolla, CA 92093-0634 (U.S.A.) Tel.(619)

534-4268;

Abbreviations: forming

unit(s);

Fax (619) 534-1422.

BHK, baby hamster ELISA,

kidney; bp, base pair(s); cfu, colony-

enzyme-linked

immunosorbent

assay;

kb, kilo-

base(s) or 1000 bp; LTR, long terminal repeat; Luc, firefly luciferase; /UC, Luc-encoding gene; neo, NPTII-encoding gene; NPTII, neomycin phosphotransferase;

nt, nucleotide(s);

now (large) fragment sarcoma virus.

ofE.

ORF, open reading

coliDNA polymerase

frame; PolIk, Kle-

I; R, purine; RSV, Rous

the overlapping pol gene is translated by ribosome frameshifting (Jacks et al., 1987) and the Epstein-Barr virus nuclear antigen-encoding mRNA, in which two nonoverlapping ORFs encoding two nuclear proteins are translated from a single mRNA (Wang et al., 1987). Two alternative mechanisms have been proposed to explain the phenomenon of tr~slational reinitiation; the scanning model (Kozak, 1989) and the internal entry model (Peabody and Berg, 1986). The scanning model postulates that after the ribosome encounters a stop codon, the 60s ribosomal subunit dissociates from the 405 subunit and the mRNA, while the 40s subunit remains associated with the mRNA and continues to scan along the RNA until it either falls off or encounters another start codon and undergoes reinitiation, The internal entry model postulates that neither subunit remains associated with the mRNA after reaching a stop codon, and that reinitiation occurs at a

168 subsequent start codon by reassociation of both ribosomal subunits. Under the scanning but not the internal initiation model, the ribosome should reinitiate primarily at the first suitable AUG codon encountered after termination at the upstream cistron. Thus, introduction of AUG codons in the intercistronic region of a dicistronic construct should

A. Two-Gene, Two-Promoter Retroviral Vector

Dicistronic

6.

Retroviral Vector

decrease expression from the downstream cistron. This has been found to be true in a number of cases, supporting ribosomal scanning as a mechanism for translational initiation (Rogers et al., 1985; Perez et al., 1987). However, there are some examples appears

to occur

in which translational

by internal

initiation

initiation

(Pelletier

and

Sonenberg, 1988; Chang et al., 1990). Initiation of translation of poliovirus mRNA has been shown to occur by internal initiation mediated by a sequence in the 5’ untranslated region (Pelletier and Sonenberg, 1988).The AUG sequences upstream from the first cistron had no effect on expression of the downstream cistron, supporting internal entry of the ribosome as a mechanism for translational reinitiation in that system. Dicistronic transcriptional units have been used in a number of gene transfer studies (Kozak, 1989). In some cases, expression from the downstream cistron was very low (Angenon et al., 1989; Balland et al., 1988). Since most of the previous studies with dicistronic transcriptional units did not characterize many of the parameters important for efficient gene expression, the low expression of downstream cistrons may have been due to unfavorable characteristics of the RNA sequence between the two cistrons. This intercistronic region, defined as extending from the stop codon of the upstream cistron to the start codon of the downstream cistron, has been shown to be critical in determining the level of expression from the downstream cistron (Kozak, 1989). Important characteristics of the intercistronic region are its length, the presence of AUG sequences and the conformity of the sequence surrounding the start codon of the downstream cistron to the eukaryotic consensus sequence (5’-GCCGCCRCCMG) in which the purine (R) at position -3 and the G immediately following the AUG start codon have been shown to be important for efficient translational initiation (Kozak, 1986). Nevertheless, dicistronic transcriptional units have been used with success in some cases (Kaufman et al., 1987; Gansbacher et al., 1990). We have been interested in defining the characteristics of dicistronic transcriptional units that are important in achieving maximal translational reinitiation and in incorporating such dicistronic units into retroviral vectors. Retroviral vectors have proven to be a versatile and effective means of transferring genes into mammalian cells (Gilboa, 1988). They can infect a wide variety of cell types with an efficiency much higher than most physicallymediated methods of gene transfer such as transfection,

Fig. 1. Structure extending

of retroviral

the 3’ LTR. A: Proviral vector LLRNL

Transcripts

form of the two-gene,

characteristics

form of the dicistronic from one another described

vectors.

are shown as arrows

from either the 5’ LTR or RSV promoter

and terminating

two-promoter

at

retroviral

of which are shown in Table I. B: Proviral

retroviral

vectors

in the characteristics

used in this study which differ of the intercistronic

region

as

in Table 1.

lipofection, or electroporation. A wide variety of different types of retroviral vectors have been developed, differing in the arrangement of introduced genes and their control elements . Most commonly, retroviral vectors contain only one gene or have two genes with one under the control of the 5’ LTR and the other under the control of an internal promoter (Fig. 1A). Single-gene vectors are less versatile than two-gene vectors because they do not allow the simultaneous transfer of a gene of interest as well as a selectable marker. Thus, vectors of the two-gene, two-promoter type are generally employed when it is necessary to select for a population ofinfected cells. While two-gene, two-promoter vectors have been very useful, they have a number of disadvantages. For some applications, such as in vitro infection and reimplantation of large numbers of primary cells, viral titers are often too low to achieve efficient infection of enough cells. This deficiency may be due in part to the phenomenon of promoter interference, in which selection for expression of one gene in a two-gene, two-promoter vector interferes with expression of the second gene (Emerman and Temin, 1986). For a retroviral vector, this can result in a decrease in the amount of full length viral transcript available for packaging. Furthermore, previous work from this laboratory (Jolly et al., 1986; Xu et al., 1989), and others (Emerman and Temin, 1984; Hoeben et al., 1991) has demonstrated that expression from integrated genes introduced by some of these vectors is often unstable. The mechanisms determining the instability are unclear, but the phenomenon of genetic or epigenetic shutdown of reporter genes varied markedly depending on a number of parameters. These included the presence or absence of selection pressure on the selectable marker and the particular genes carried by the retrovirus, suggesting that internal sequences predispose in different ways to instability.

169 To address the above problems, we have constructed and begun to characterize dicistronic retroviral vectors. As opposed to two-gene, two-promoter retroviral vectors which produce two overlapping transcripts (Fig. lA), dicistronic vectors produce only one transcript from which both cistrons are translated (Fig. 1B). Translation of the downstream cistron depends on the ability of the ribosome to reinitiate at internal AUG sequences (Kozak, 1986). Our results show that expression of the downstream cistron is very efficient when intercistronic AUG sequences are removed, an intercistronic distance of 15-78, bp is maintained, and the sequence surrounding the start codon of the downstream cistron conforms to the eukaryotic consensus sequence. In these studies, dicistronic vectors had significant advantages over a two-gene, two-promoter retroviral vector, including a smaller genome, higher viral titer and increased stability of gene expression.

distance of 84 bp and contains three intercistronic ATG sequences derived from the Zuc3’ untranslated region. The sequence surrounding the neo start codon does not conform to the eukaryotic consensus sequence. Since the only promoter in this vector is located in the 5’ LTR, expression of the neo downstream gene depends on translational reinitiation. Transfection of this plasmid into BHK cells resulted in very few G418-resistant colonies (Fig. 2) and produced less than one infectious virus particle per ml (data not shown). Since we believed that the most likely

has been altered

by in vitro mutagenesis

translation-initiation Plasmid

pLLl5NL

ated in vitro mutagenesis

Plasmid

I

pLL78NL

of retroval

was constructed

fragment

to introduce

vectors

et al., 1989) so that

and a 15-bp intercistronic

by digesting

with PolIk

pLL148NL

(Sambrook

from pBluescript

any ATG sequences was constructed

1987).

region derived from the 3’ end

BglII site remained.

larly filled in and the two fragments

The product

of the

(data not shown).

pLLl5NL

with BglII

et al., 1989). A 59-bp

SK - (Stratagene

Inc.) was simi-

were ligated. This was designed into the intercistronic

by digesting

pLL78NL

pBluescript lowed

SK -

which

was created

by filling in with T4 DNA

with BamHI

and from

by Hind111 + Sac1 digestion

polymerase.

not

region. Plasmid

filling in the ends with PolIk. It was ligated with a 66-bp fragment The orientation

fol-

of the

fragment

is with the Hind111 site 5’ to the Sac1 site. This does not

introduce

any intercistronic

constructed

Characteristics

(Sambrook

was verified by dideoxy DNA sequencing

HincII-XbaI

TABLE

of pLL71NL

an introduced

to the eukaryotic

and Capecchi,

by oligodeoxyribonucleotide-medi-

in the intercistronic

and filling in the ends

(a) Structure of retroviral constructs Our prototype vectors contain the luc reporter gene and the neo selectable marker gene. They have the organization 5’ LTR-luc-neo-3’ LTR (Fig. 1B). The first dicistronic vector constructed, pLL84NL (Table I), has an intercistronic

(Thomas

of luc (de Wet et al., 1987) were eliminated mutagenesis

AND DISCUSSION

to conform

sequence

was constructed

three ATG sequences region containing

RESULTS

consensus

in a manner

ATG sequences. exactly

analogous

Plasmid

pLL218NL

was

to pLL148NL.

b NA, not applicable. ‘-’ luc stop codon (TAA) and neo start codon (ATG) are in bold italics.

Plasmid”

Intercistronic

Optimized

Intercistronic

Intercistronic

distance

start codon’

ATGs’

’ Intercistronic

(bp)

neo

&&

sequences

sequence:

are underlined.

5’-TAAAmTAACTGTATTCAGCGm

GACGAAATTCTTAGCTATTGTA~GGGGATCTGATCAAGpLLRNL

NAb

no

NA

AGACAGGATCAGGATCGTITCGCATGA.

pLL84NL

84’

no

3

d Intercistronic

pLL71NL

71d

yes

3

~ACGAAATICTTAGCTAl-TGTA~GGGGATCCCCCGGGCT-

sequence:

5’-TAAAmTAACTGTATTCAGCGu

pLL15NL

15’

yes

0

GCAGCCAATATGG.

pLL78NL

7gf

yes

0

e Intercistronic

sequence:

5’-TAAAGATCTGCAGCCATTATGG.

pLL148NL

148a

yes

0

f Intercistronic

sequence:

5’-TAAAGATCCTAGAACTAGTGGATC-

pLL218NL

218”

0

CCCCGGGCTGCAGGAATTCGATATCAAGCTTATCGATACC-

“The

letter p in front of the name of a retroviral

vector

designates

plasmid construct. The absence of a p signifies that the plasmid used to produce an infectious retrovirus. pLLRNL is a two-gene, moter retroviral (pLLBNL)

while luc (pL_RNL)

gene is driven by the retroviral

(Xu et al., 1989). All other constructs

Plasmid

was constructed

pLL84NL containing

the firefly luciferase

Plasmid

pLL71NL

are dicistronic.

by ligating a 1.7-kb BarnHI-Hind111 gene (de Wet et al., 1987) into

pLRbRNL (Huang et al., 1988) which had been Bg1II + Hind111 to remove the retinoblastoma encoding RSV promoter.

has been two-pro-

vector in which neo (pLLR&JL) gene is driven by the RSV

promoter

5’ LTR (pLLRNL) fragment

a

was constructed

end of neo from pLL84NL

(originally

and Berg, 1982) extending

to the internal

digested sequence

by replacing

derived from pSV2Neo)

with and the 5’

(Southern

SphI site with a neo 5’ fragment

GTCGATCTGCAGCCAATATGG. a Intercistronic sequence: 5’-TAAAGATCCTAGAACTAGTGGATCAGCTTGATATCGAATTCCTGCAGCCCGGGGGATCCACTAGTTCTAGAGCGGCCGCCACCGCGGTGGGATCCCCCGGGCTGCAGGAATTCGATATCAAGCTTATCGATACCGTCGATCTGCAGCCAATATGG. h Intercistronic

sequence:

5’-TAAAGATCCTAGAACTAGTGGATC-

AGCTTGATATCGAATTCCTGCAGCCCGGGGAGCTTGATATCGAATTCCTGCAGCCCGGGGGATCCACTAGTTCTAGAGCGGCCGCCACCGCGGTGGGATCCACTAGTTCTAGAGCGGCCGCCACCGCGGTGGGATCCCCCGGGCTGCAGGAATTCGATATCAAGCTTATCGATACCGTCGATCTGCAGCCAATATGG.

’ This refers to the conformity codon mG).

with the eukaryotic

of the sequence

consensus

sequence

surrounding

the neo start

(5’-GCCGCCRCC-

derived contain

from pMClNeo (Stratagene, Inc.). This fragment does not the point mutation in neo described by Yenofsky et al. (1990).

’ This refers to the presence or absence of three ATG sequences derived

Plasmid

pMClNeo

from the 3’ end of the /UC gene, in the luc-neo intercistronic

differs from pSV2Neo

in that the pMClNeo

neo gene

region.

170

Fig. 2. Colony formation

mediated

by pLLRNL

media plus 10% fetal calf serum and in log-phase (Graham

cells were fixed and stained with Giemsa

per ng transfected

plasmid

retroviral

growth were transfected

and Van der Eb, 1973) for 6 h and placed in medium

stably transfected of colonies

and dicistronic

containing

vector plasmids.The with 2.5 pg of plasmid

5 x lo4 BHK cells grown in Dulbecco’s

400 ,ag/ml G418 two days after transfection.

stain six days after being placed in selection.

DNA from two independent

experiments,

reason for this was poor neo expression, we examined the characteristics of dicistronic vectors which are important for expression of the downstream cistron. Accordingly, a series of dicistronic vectors differing in various characteristics of the intercistronic region were constructed (Table I). The parameters studied were intercistronic distance, the presence of intercistronic AUG start codons, and the sequence surrounding the downstream neo start codon. Expression of both the luc and neo cistrons in the dicistronic constructs was compared to the two-gene, twopromoter retroviral vector pLLRNL, in which neo is driven by a strong internal promoter consisting of an RSV LTR sequence from which the polyadenylation sequence has been deleted (Xu et al., 1989)(Fig. 1A). As expected, Northern-blot analysis of cells infected with LLRNL and probed with a neo fragment showed two bands, one from the 5’ LTR and the other from the internal RSV promoter while cells infected with the dicistronic vector LL15NL had a single band from the 5’ LTR (data not shown). In order to measure the efficiency oftranslational reinitiation, two separate assays for neo expression were performed. In the first, the plasmid constructs listed in Table I were transfected into BHK cells (Littlefield and Basilica, 1966) and assayed for their ability to form colonies under selection with G418 (Fig. 2). The ability of transfected cells to grow and form colonies in the presence of G418 is dependent on the level of NPTII enzyme activity. We also used an NPTII ELISA assay to provide an independent measure of lzeo expression (Fig. 3). Retroviral plasmid constructs were transfected into BHK cells and cell extracts were prepared for NPTII ELISA assay after 48 h . Since Iuc is driven by the same 5’ LTR promoter in all of

the results

modified

DNA (see Table I) by calcium phosphate Numbers

Resultant

in parentheses

of one of which are displayed

Eagle’s

coprecipitation

colonies

representing

are the average

number

here.

the constructs, Luc activity was used as a measure of transfection efficiency. (b) Effect of the sequence surrounding the neo start codon on translational reinitiation The importance of adherence to the eukaryotic consensus sequence surrounding the downstream start codon can

pLLRNL

Fig. 3. The NPTII

ELISA

1

pLL78NL

pLL71NL pLL84NL

pLLlSNL

assay.The

pLLZl4NL pLL148NL

2.5 x lo5 BHK

cells were trans-

fected with 10 pg of plasmid DNA (see Table I) by calcium phosphate coprecipitation (see Fig. 2 legend). At two days after transfection, cellular extracts were prepared by washing with PBS (137 mM NaCh2.7 mM KC1/4.3 mM NazHP0,.7H,0/1.4 mM KH,PO,) , scraping the cells into 1 ml of PBS in a 1.5 ml microcentrifuge 3 cycles of freeze-thawing.

This extract

tube and subjecting was centrifuged

the cells to

for 10 min at 10

000 x g and the supernatant was used for both NPTII ELISA and Luc assays.The NPTII ELISA assays were performed using an NPTII ELISA assay kit as described by the manufacturer (5’-3’, Inc., West Chester, PA). Luc assays were performed as previously described (de Wet et al., 1987).The NPTII levels are expressed The standard error bars.

deviations

as pg NPTII protein/lo6

of the mean of three experiments

light units.

are shown as

171 amount

of NPTII.

Intercistronic

distances

greater than 78

be determined by a comparison of the amount of neo expression from LL71NL and LL84NL. In the colony

bp resulted

assay, pLL84NL, which has three intercistronic ATG sequences and an unfavorable neo start codon (Table I), produced very few colonies. Transfection with pLL7 lNL, which has a similar intercistronic distance and the same

produced, with transfection of both pLL148NL and pLL2 18NL resulting in relatively poor NPTII production. The reason for the discrepancy between our results and those of Kozak may have to do with the specific sequences

intercistronic ATG sequences but has a neo start codon conforming to the eukaryotic consensus sequence, produced three times as many colonies (Fig. 2). In the NPTII ELISA assay, LL71NL expressed seven times as much NPTII as LL84NL, documenting its importance in ob-

of the intercistronic region, which are completely different from each other except for the intentional avoidance of intercistronic ATG sequences. As a consequence of the way

taining optimal translational reinitiation. We do not believe that the difference in intercistronic distance accounts for the difference in neo expression between LL7 1NL and LL84NL since LLl5NL and LL78NL, which differ only in intercistronic distance, have the same neo expression (see section c below).

cistronic region. These may contribute to RNA secondary structure which could decrease the ability of the ribosome to scan and reinitiate translation.

(c) Effect of intercistronic ATG sequences on neo expression The significance of intercistronic ATG sequences can be seen by comparing the number of colonies produced by pLL71NL and pLL78NL. These plasmids differ primarily in the presence of intercistronic ATG sequences, with pLL71NL having three out-of-phase intercistronic ATG sequences while pLL78NL has none. One of the intercistronic ATG sequences has a surrounding sequence that matches the eukaryotic consensus sequence very closely (Table I) (de Wet et al., 1987). Ifthe 40s ribosomal subunit reinitiates by a scanning mechanism, intercistronic AUG sequences would be expected to interfere with expression of the downstream cistron. Our results are consistent with this prediction and therefore support the scanning model for translational reinitiation in this case.

(d) Effect of intercistronic distance on neo expression Intercistronic distance proved to be an important variable in the determination of optimal expression of the downstream neo cistron. Kozak (1987) found that an intercistronic distance of between 75 and 150 bp allowed optimal expression of the downstream cistron, whereas a short distance of 10 bp was deleterious. She speculated that this effect of very short distances was due to the possibility that binding of the 40s ribosomal subunit to initiation factors required more time than was permitted by scanning over a short distance. We did not find that a short intercistronic distance of 15 bp was deleterious to downstream cistron expression. Transfection of pLLlSNL, with an intercistronic distance of 15 bp, into BHK cells, resulted in approximately the same number of colonies and neo expression as pLLRNL. Transfection of pLL78NL, which differs from pLL15NL only in intercistronic distance, also resulted in approximately the same number of colonies and

that they pLL218NL

in a substantial

drop in the amount

of NPTII

were constructed, both pLL148NL and contain inverted repeat sequences in the inter-

(e) Virus titers Under most conditions, retroviral vectors provide a maximum titer of approx. lo6 infectious particles/ml virus supernatant (Miller, 1990). Additionally, such titers can only be achieved by screening a number of clones and selecting for the best virus producers. Although such a titer is adequate for many gene transfer studies, it can be a serious limitation for some gene transfer experiments where large numbers of primary cells must be infected in order to achieve a meaningful biological effect or where direct in vivo delivery of vector is desirable. Therefore, we compared the virus titers resulting from infection of a retroviral packaging cell line with LLRNL and LLlSNL. The PA3 17 retroviral packaging cells were transfected with 20 pg of pLL 15NL or pLLRNL and supernatant was harvested after two days and used to infect $2 packaging cells. Hundreds of independently infected colonies were pooled after one week of selection with G418 so that effects of retroviral integration site on virus titer were averaged out. Although this procedure results in lower titers than the maximal obtainable with individual clones, we believe that it provides a more valid comparison of the relative titers. Virus-containing supernatant from such bulk-infected cultures was harvested and titered by infection of rat 208F cells (Quade, 1979) with dilutions of supernatant from a near confluent culture of producer cells in the presence of 8 pg Polybrene/ml (Sigma). Infected cells were placed in selection with 400 pg G4 18/ml one day after infection and stained with Giemsa after seven to ten days to count the G418-resistant colonies. Supernatant from $2 cells infected with LLRNL virus had a titer of 1.7 x lo4 k 1.5 x lo3 cfu/ml while that from $2 cells infected with LL15NL had a titer of 7.2 x lo4 & 6.4 x lo3 cfu/ml. Thus, deletion of the internal promoter resulted in a substantial increase in the virus titer. As the factors that affect titer are complex, the reasons for the higher titer with the dicistronic vector are not clear. Possibilities include a difference in packaging efficiency because of the different size of the

172

viral RNAs or elimination of the potential for promoter interference by removal of the internal promoter (Emerman and Temin, 1986). Further studies would be required to distinguish between these alternatives. (f) Stability

of gene expression

One of the purposes in undertaking these studies was to develop retroviral vectors that allow increased stability of expression of integrated genes. Previous studies have demonstrated that instability of integrated genes introduced by retroviral vectors with internal promoters can be a serious problem (Emerman and Temin, 1984; Xu et al., 1989; Hoeben et al., 1991). Many different mechanisms have been found to be responsible for this instability, including point mutations, deletions and methylation. The two-gene, two-promoter vector used in this study, LLRNL, has been shown to lose Luc activity by multiple mechanisms, including frequent deletions (Xu et al., 1989). We hypothesized that simplification of the structure of the vector through the elimination of the internal promoter and the resulting tighter linkage between the expression of the two cistrons should lead to increased stability of gene expression. Therefore, we examined the stability of Zuc expression from cells infected with the retroviral vectors LLRNL and LL15NL in the presence and absence of G4 18 selection (Fig. 4). Supernatant from $2 cells producing either LLRNL or LLl5NL was used to infect lo5 rat 208F cells. Infected cells were placed in G4 18 selective medium

I

-01

0

60 20 40 Days post infection

Fig. 4. Stability ofgene expression viral packaging

from LLRNL and LLISNL.

cell lines $2 (ecotropic)

The retro-

and PA317 (amphotropic)

have

been previously described (Miller, 1990). All cells were maintained in Dulbecco’s modified Eagle’s medium plus 10% fetal calf serum (Gibco). The

PA317

was

maintained

Szybalski,

1962) to prevent

phenotype

associated

in the reversion

HAT

medium

with loss of the introduced

enr genes. Cells were transfected

(Szybalska

and

to the thymidine-kinase-negative

by calcium

retroviral

phosphate

gag,

pol and

co-precipitation.

The Luc activity was measured as light units per pg protein. A, LLRNL +G418; A,LLRNL-G418;O,LLlSNL +G418;m,LLlSNL-G418. Each point represents the average from three experiments. The average standard

error from these experiments

for LLRNL -G418.

-G418,

10% for LLl5NL

is 11 y0 for LLRNL +G418,

+ G418,27%

and 13% for LLlSNL

24 h after infection, and the hundreds of colonies resulting

from each infection were pooled after one week of selection. Three sets of pooled clones were split into media with and without G4 18. Cell extracts were prepared every seven to ten days for Luc and protein assays. Strikingly, Luc activity from cells infected with LLRNL in the absence of G418 decreased markedly and rapidly during the subsequent weeks. In contrast, cells infected with LLl5NL showed stable and even increasing Luc activity under identical conditions. In the presence of G418, both LLRNL and LLl5NL showed an initial increase in Luc activity. This is most likely due to a growth advantage of clones which integrated into genomic sites favorable for high level transcription and which consequently had increased neo and luc expression. The reason for the increase in Luc activity over time in cells infected with LL 15NL and not maintained in G4 18 is not clear. Regardless, this experiment indicates an improvement in the stability of expression of integrated genes transferred by this dicistronic vector over its comparable two-gene, twopromoter vector in the absence of selection and demonstrates that the internal promoter itself may lead to instability. This finding is not limited to the use of RSV as an internal promoter as Emerman and Temin (1984) observed that the a-globin promoter caused genetic instability when used as an internal promoter in a retroviral vector. Additionally, although the internal RSV promoter is also derived from a retroviral LTR, it does not have significant sequence homology to the Moloney murine leukemia virus LTR used here (Van Beveren et al., 1985), so deletion of luc by homologous recombination between the LTRs and the RSV promoter is unlikely. However, since some vectors containing internal promoters appear to be relatively stable (Xu et al., 1989) the characteristics of the promoter or other elements of the retroviral vector contributing to genetic instability are not understood. Internal transcriptional units exist in introns of a number of genes of higher eukaryotes, demonstrating that this need not be an inherently unstable genetic organization (Levinson et al., 1990). Expression of the two genes in a dicistronic retroviral vector is more tightly linked than in a two-promoter vector because both are transcribed from the same promoter and translation of the downstream cistron is dependent on translation of the upstream cistron. For example, in a two-promoter vector such as LLRNL, a mutation or epigenetic event affecting the promoter function ofthe 5’ LTR would only affect expression of the upstream gene, leaving expression of the selectable marker intact. A similar event in a dicistronic vector would affect expression of both genes. In the presence of selection, we expected that this tighter linkage might manifest as increased stability of fuc expression in cells infected with the LL15NL compared to LLRNL. However, no difference was seen in the presence

173 of G418. This may be because is fairly stable in the presence

/UCexpression from LLRNL of G418 (Xu et al., 1989). It

is possible that examination over a longer time would reveal a decrease in Luc activity from LLRNL compared LL15NL in the presence of G418. The finding of unstable expression from two-promoter vectors in the absence of selection has implications that are often disregarded in the design or interpretation of gene therapy experiments, both in vitro and in vivo. The design of the stability experiment reported here mimics the most commonly used model for gene therapy, i.e., cells are infected in bulk in vitro, selected for a period of time, then removed from selection prior to transplantation back to a recipient problems

animal or human patient. One of the major with model systems used so far has been

obtaining prolonged in vivo expression from a transgene in a grafted cell or tissue. Many studies have demonstrated that while expression is high at early times following transplantation, it often decreases rapidly (St. Louis and Verma, 1988; Wilson et al., 1990) Although there are many possible explanations for this phenomenon, genetic instability is almost certainly a factor in some cases. Thus, if shown to be a generalizable feature of dicistronic retroviral vectors, the increased stability of the dicistronic vectors examined in this study may prove to be of significant value as retroviral vectors come into wider use in in vivo applications of gene transfer.

marker

may be vital in order to achieving

balanced

biological

regulated

and

activity.

ACKNOWLEDGEMENTS

We thank Kathy Bouic for help with DNA sequencing and Klaus Romer and Paul Johnson for helpful comments on the manuscript. This work was supported by a grant from the Stern Foundation and by NIH grants GM 13538 to F.L. and CHD 20034 and CA 51495 to T.F.

REFERENCES Angenon,

G., Uotila, J., Kurkela,

Montagu,

M. and

scriptional

units

S.A., Teeri, T.H., Botterman,

Depicker,

A.: Expression

in transgenic

tobacco.

J., Van

of dicistronic

tran-

Mol. Cell. Biol. 9 (1989)

5676-5684. Balland,

A., Faure,

T., Carvallo,

D., Cordier,

P., Ulrich, P., Fournet,

de La Salle, H. and Lecocq, J.-P.: Characterization processed

forms

of human

CHO cells transformed

recombinant

factor

with a polycistronic

B.,

of two differently IX synthesized

in

vector. Eur. J. Biochem.

172 (1988) 565-572. Chang,

L.-J., Ganem,

the hepadnaviral

D. and Varmus, polymerase

H.E.: Mechanism

of translation

of

(P) gene. Proc. Nat]. Acad. Sci. USA 87

(1990) 5158-5162. de Wet, J., Wood, K.V., Deluca, Firefly luciferase

M., Hehnski,

gene: structure

D.R. and Subramani,

and expression

in mammalian

S.: cells.

Mol. Cell. Biol. 7 (1987) 725-737.

(g) Conclusions (I) We have studied the characteristics of dicistronic transcriptional units that allow efficient translational reinitiation and have shown that a luc-neo dicistronic transcriptional unit incorporated into a retroviral vector expresses the downstream neo cistron at high levels. The dicistronic retroviral vectors examined here had a number of advantages over the corresponding two-gene, two-promoter retroviral vector including significantly higher titers, smaller genome size allowing for insertion of larger genes, and improved stability of gene expression. (2) The successful development of dicistronic retroviral vectors raises the possibility of designing vectors expressing more than two genes. Although retroviral vectors containing two or more genes with each under the control of a separate internal promoter have been developed (Emerman and Temin, 1984; Overell et al. 1988), this approach has proven to be difficult, due partly to significant genetic instability of such vectors (Emerman and Temin, 1984). The efficient translational reinitiation seen in the vectors used here suggests that it may be possible to create vectors containing three or more genes in tandem with no internal promoters. For many systems, such as multisubunit proteins or metabolic pathways with multiple steps, transfer of several genes in addition to a selectable

Emerman,

M. and Temin,

retrovirus

vectors

H.M.: High-frequency

containing

exogenous

deletion

DNA

in recovered

with promoters.

.I.

Virol. 50 (1984) 42-49. Emerman,

M. and Temin, H.M.: Comparison

avian and murine

retrovirus

vectors.

of promoter

Nucleic

suppression

in

Acids Res. 14 (1986)

9381-9396. Gansbacher,

B., Zier,

Gilboa,

K., Daniels,

E.: Interleukin

tumorigenicity

B., Cronin,

2 gene transfer

and induces

protective

K., Bannerji,

into tumor immunity.

R. and

cells abrogates

J. Exp. Med. 172

(1990) 1217-1224. Gilboa, E.: Retroviral

gene transfer:

applications

to human therapy.

Adv.

Exp. Med. Biol. 241 (1988) 29-33. Graham,

F.L. and Van der Eb, A.J.: A new technique

infectivity Hoeben,

of human

adenovirus

R.C., Michielsen,

H. and

Van

der

for the assay of

5 DNA. Virology 52 (1973) 456-467.

A.A.J., Van der Jagt, R.C.M., Van Ormondt,

Eb, A.J.:

Inactivation

of the Moloney

murine

leukemia virus long terminal repeat in murine fibroblast cell lines is associated with methylation and dependent on its chromosomal position. J. Virol. 65 (1991) 904-912. Huang,

H.-J.,

Yee,

Friedmann, neoplastic

phenotype

ceils. Science Jacks,

J.-K.,

Shew,

T., Lee, E.Y.-H.P.

J.-Y.,

Chen,

P.L.,

and Lee, W.-H.:

by replacement

Bookstein,

Suppression

R., of the

of the RB gene in human cancer

242 (1988) 1563-1566.

T., Townsley,

K., Varmus,

ribosomal

frameshifting

mammary

tumor

H.E. and Majors,

events are required

virus gag-related

J.: Two efficient

for synthesis

polyproteins.

of mouse

Proc. Natl. Acad.

Sci. USA 84 (1987) 4298-4302. Jolly, D.J., Willis, R.C. and Friedmann,

T.: Variable

able provirus after retroviral vector gene transfer Mol. Cell. Biol. 6 (1986) 1141-l 147.

stability

of a select-

into human

cells.

174 Kaufman,

R.J., Murtha,

polycistronic

P. and Davies,

mRNAs

M.V.: Translational

and their utilization

to express

efficiency of heterologous

that modulates

translation

by eukaryotic

ribosomes.

Cell 44

(1986) 283-292. Kozak,

M.: Effects ofintercistronic

by eukaryotic Kozak,

ribosomes.

M.: The scanning

length on the efficiency of reinitiation an update.

J. Cell Biol.

J.: A transcribed Genomics Littlefield,

S., Lakich,

D., Hammonds

gene in an intron

of the human

J.W. and Basilica,

C.: Infection

virus. Nature

VIII gene.

packaging

K.E. and Cosman,

promoter

retroviral

of thymidine-kinase

deficient

211 (1966) 250-252.

Miller, D.A.: Retrovirus

mammalian

D.: Stably transmitted

vectors and their use in transformation

triple-

ofprimary

of mammalian

cell

mRNAs.

occurs

Mol.

Cell.

in the trans-

Biol.

6 (1986)

J. and

Sonenberg,

mRNA

N.: Internal

directed

initiation

by a sequence

of translation

derived

of

from poliovirus

334 (1988) 320-325.

Perez, L., Wills, J.W. and Hunter, effects of upstream

E.: Expression

initiation

K.: Transformation

codons.

of the Rous sarcoma replacement

vector:

J. Virol. 61 (1987) 1276-1281.

of mammalian

virus and avian erythroblastosis

cells by avian virus. Virology

myelocyto98 (1979)

of human

cell

cell lines, IV.

of a biochemical

M.R.: Site-directed

in mouse embryo-derived with functional

N., Varmus,

trait. Proc.

S.: Nucleotide

and structural

Laboratory,

mutagenesis

by gene

stem cells. Cell 5 1(1987) 503-5 12. analysis.

H. and Coffin, J. (Eds.),

Harbor

Cold

sequences

RNA Tumor

Spring

comple-

In: Weiss, R., Teich,

Harbor,

Viruses. NY,

Cold

1985, pp.

148.

F., Petti, L., Braun,

Epstein-Barr infected,

D., Seung,

virus mRNA

encodes

growth-transformed

S.G., Fraley, R.T., Horsch,

L.A., Fink, CL., dence for ribosome in transformed

Wilson, J.M., Chowdhury, A., Mulligan, transplanted

S. and Kieff, E.: A bicistronic two nuclear

lymphocytes.

proteins J.

in latently

Virol.

61

(1987)

Mozer,

R.B., Levine, A.D., Flick, J.S., Brand,

T., O’Connell,

scanning

K. and Sanders,

during translation

initiation

P.R.: Eviof mRNAs

plant cells. Plant Mol. Biol. Rept. 3 (1985) 11 l-1 16.

N.R., Grossman,

R.C. and Chowdhury, in low density with genetically

M., Waksman,

J.R.: Temporary

lipoprotein

R., Epstein,

amelioration

receptor-deficient

modified hepatocytes.

of

rabbits

Proc. Natl. Acad.

Sci. USA 87 (1990) 8437-8441. Xu, L., Yee, J.-K., long-term Virology Yenofsky,

461-465. Rogers,

W.: Genetics

transformation

K.R. and Capecchi,

Spring

to somatic

Sci. USA 48 (1962) 2026-2034.

hyperlipidemia

virus erzv gene from a simian virus 40 late-region

matosis

heritable

approach

945-954.

RNA. Nature

Quade,

cells to an-

Sci. USA 85 (1988) 3150-3154.

Van Beveren, C., Coffin, J. and Hughes,

Wang,

2695-2703. eukaryotic

Thomas,

567-l

cells. Mol. Cell. Biol. 8 (1988) 1803-1808.

of mammalian

I.M.: An alternative

Proc. Natl. Acad.

E.H. and Szybalski,

mented

cells. Hum. Gene Ther. 1 (1990) 5-14.

D. and Berg, P.: Termination-reinitiation

Pelletier,

Szybalska,

targeting

Overell, R.W., Weisser,

lation

factor

A

Press,

tibiotic resistance with a bacterial gene under control of the SV40 early region promoter. J. Mol. Appl. Genet. 1 (1982) 327-341.

Natl. Acad.

7 (1990) l-11.

cells with polyoma

Peabody,

Jr., G. and Gitschier,

Cloning.

Laboratory

NY, 1989.

P.J. and Berg, P.: Transformation

DNA-mediated

B., Kenwrick,

T.: Molecular

2nd ed. Cold Spring Harbor

Cold Spring Harbor, Southern,

gene therapy.

108 (1989) 229-241. Levinson,

E.F. and Maniatis,

Manual,

St. Louis, D. and Verma,

Mol. Cell. Biol. 7 (1987) 3438-3445.

model for translation:

J., Fritsch,

Laboratory

genes in mammalian cells. EMBO J. 6 (1987) 187-193. Kozak, M.: Point mutations define a sequence flanking the AUG initiator codon

Sambrook,

stability

of moloney

murine leukemia

R.L., Fine, M. and Pellow, J.W.: A mutant

selection

3435-3439.

T.: Factors

affecting

virus-based

vectors.

171 (1989) 331-341.

transferase tic

Wolff, J.A. and Friedmann,

II gene reduces the resistance pressure.

Proc.

Nat].

neomycin

oftransformants

Acad.

Sci. USA

phosphoto antibio87 (1990)