Importance of 3' non-coding sequences for efficient

Journal o f General Virology (1993), 74, 39-46.

Printed in Great Britain

39

Importance of 3' non-coding sequences for efficient retrovirus-mediated gene transfer in avian cells revealed by self-inactivating vectors Frederic F l a m a n t , 1. D e n i s e Aubert, 1 Claude Legrand, 1 Francois-Loi'c C o s s e t 2 and J a c q u e s S a m a r u t I Laboratoire de Biologic MoldcuIaire et Cellulaire, I N R A - C N R S UMR 49, Ecole Normale Sup&ieure de Lyon, 46 allde d'Italie, 69364 Lyon Cedex 07 and 2Laboratoire de Biologic Cellulaire, Universitd Claude Bernard Lyon I, 43 boulevard du 11 novembre 1918, 69622 Villeurbanne Cedex, France

Avian leukosis virus-derived vectors were constructed with an internal transcriptional promoter and various 3' non-coding sequences. Deletions were introduced into the downstream U3 long terminal repeat (LTR) to obtain self-inactivation of LTR-mediated transcription after one round of replication. However, 3' non-coding

sequences appeared to determine not only selfinactivation of the vectors but also gene transfer efficiency. Further analysis revealed the influence of these sequences on both internal gene expression and RNA packaging. One construct permitted gene transfer while inactivating 5' LTR-promoted transcription.

Introduction

are unpredictable (Iwasaki & Temin, 1990; Yee et al., 1987; Yu et al., 1986; Olson et al., 1992). Here we describe the construction of several avian leukosis virus-derived vectors with various modified 3' non-coding sequences. The influence of these sequences on virus replication is further analysed. DNA transfection experiments and RNA analysis revealed that the efficiency of virus-borne gene expression and viral RNA packaging are dependent on 3' non-coding sequences.

Retrovirus vectors provide an efficient method for the introduction of foreign genes into various cell types, where they usually ensure a high level of stable gene expression (Eglitis & Anderson, 1988). Genes of interest can be expressed either from the 5' long terminal repeat (LTR) of the retrovirus or from an internal promoter. In the latter case, however, the proximity of viral transcriptional signals can impair the activity of the internal promoter (Emerman & Temin, 1984, 1986). An elegant way to overcome this problem is to design a vector where at least some of the viral transcription signals are lost after one round of replication. Such vectors, called self-inactivating vectors, have been previously derived from both murine leukaemia viruses and avian spleen necrosis virus (SNV) by mutating the 3' LTR; in the transfected helper cells, this mutation is not expected to impair virus production, but after one round of virus replication both LTRs are mutated and the 5' LTR-driven transcription is reduced or abolished (Cone et al., 1987; Dougherty & Temin, 1987; Soriano et al., 1991; Yu et al., 1986). This inactivation process provides an important improvement in the safety of in vivo gene transfer. The ability of self-inactivating vectors to produce high titre recombinant viruses and to inactivate the LTR promoter relies mostly on the mutations that are created. However, the consequences of mutating the 3' noncoding viral sequences on RNA transcription, processing, translation, packaging and reverse transcription 0001-1140 © 1993 SGM

Methods Plasmid construction (i) O V A backbone. OVA was constructed by assembling in the polylinker of pBluescript (Stratagene) (5' to 3'): a fully active hybrid LTR [referred to as the RSV-D LTR in the text (RSV, Rous sarcoma virus)], an RAV(1) leader sequence [NdeI-EcoRI fragment from pRSVneo (Gorman et aI., 1982) and EcoRI Xhol from pRAV(1) (Payne et al., 1981)], the Neo a coding sequence fused to the gag sequence (XhoI-XbaI fragment ofpTXN5"; Benchaibi et al., 1989) and the simian virus 40 (SV40) promoter followed by a mutated version of the Escherichia coli lacZ gene, coding for a fl-galactosidase targeted to the cell nucleus (SalI-BamH1 fragment from pMuMLVSVnlslacZ; Bonnerot et al., 1987). Due to the insertion of small pBluescript polylinker fragments, this structure is flanked by two Spel restriction sites. This SpeI fragment was inserted upstream of the various LTRs described in the following section, in some cases after filling in of the protruding ends with Klenow polymerase. The last letters of OVA vectors' names denote the origin of the 3' LTR. (ii) 3 ~ Non-coding sequences o f 0 VA-derived vectors. The OVA-D 3' LTR was derived from the NdeI Xhol fragment of OVA which was cloned between the SmaI and XhoI sites of pBluescript. The OVA-Rdeleted 3' LTR was created by assembling the 3' part of the RAV(2) genome (from XhoI to a ClaI linker inserted in the SphI site which is

40

F. Flamant and others'

located 149 nucleotides upstream of the cap site) and the 5' part of the RAV(1) LTR (from a ClaI linker inserted into the TaqI site 15 nucleotides upstream of the cap site). This deletion removes the TATA and CCAAT boxes but leaves the enhancer element intact (Aubert et al., 1991). This LTR is referred to as the RAV(1)-deleted LTR in the text. In OVA-ZZ the 3' LTR comes from the H i n d I I I - B a m H I fragment of pGJ18 (Cullen et al., 1983) carrying two copies of the RAV(0) LTR. OVA-Zdel was obtained by deleting the tandem repeats of RAV(0) LTRs by digesting with NdeI and performing a ligation at low DNA concentration. This leaves a single copy of the RAV(0) LTR deleted from nucleotide - 9 8 to nucleotide - 5 9 (from the cap site). OVA-Zcrip was similar to OVA-Zdel but the TATA box was deleted further by site-directed mutagenesis (Muta-Gene kit, Bio-Rad). The sequence of the mutagenic oligonucleotide was 5' ACGAAGCTATGTAGTAAGCTGTTGCCACC 3'. The deletion covers nucleotides - 2 7 and - 3 2 and creates a SnaBI restriction site. The structure of this LTR was confirmed by DNA sequencing. Cell culture. All reagents were purchased from Gibco. Chicken embryonic fibroblasts (CEFs) were prepared from C/O SPAFAS 10day-old embryos. CEF, QT6 (Moscovici et al., 1977) and Isolde helper cells (Cosset et al., 1990) were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 5 % foetal calf serum, 5 % newborn calf serum, 1% chicken serum, 2 % tryptose phosphate broth, penicillin (1000 units/ml), and streptomycin (0.05%). Medium was changed every 3 days. Phleomycin (50 gg/ml) and hygromycin (50 lag/ml) were added to maintain the virus production of the Isolde cells. D N A transfection, virus production and titration. Plasmid DNA purified by CsC1 banding (Maniatis et al., 1982) was used to transfect Isolde helper cells using either calcium phosphate precipitation (Maniatis et al., 1982), polybrene (Kawai & Nishizawa, 1984), DOTAP (Boehringer-Mannheim) or electroporation (10 gg/ml of DNA in DMEM, 250 gF, 350 V with a Bio-Rad Gene-Pulser). Cells were selected with G418 (500 ~tg/ml, Gibco), starting 24 h after transfection, for at least 10 days. Virus supernatants were harvested from subconfluent cells 16 h after medium change, and centrifuged (4000 g, 5 rain) to remove cell debris. For titration, 50 to 500 gl of helper cell supernatant was spread on QT6 cells or CEFs (5 x 105 cells per 25 cm 2 dish). Cells were selected with G418 24 h after infection, or stained for fl-galactosidase activity 48 h after infection (Sanes et al., 1986). To ensure that the virus preparations were free of replication-competent virus, infected cells resistant to G418 were grown for another week and assayed in situ for the presence of replicating virus by using an anti-p27 antibody (Savatier et al., 1989). R N A and D N A analysis. Virion RNA was extracted from centrifuged supernatant (45 min at 35000 r.p.m, in an SW41 Beckman rotor). Cellular RNA and DNA were purified after proteinase K digestion and phenol-CHCl~ extraction as described (Maniatis et al., 1982). Southern blots (cellular DNA), Northern blots (cellular RNA) and slot blots (virion RNA) were hybridized with an in vitro synthesized RNA probe labelled with 32p using standard techniques. Quantification was performed by cutting out nylon membrane pieces, immersing in a liquid scintillation cocktail (Beckman) and measuring radioactivity.

Results

Construction of a highly efficient vector derived from R S V with an internal SVdO promoter As a first step toward the construction of a selfinactivating vector, we needed to construct an efficient vector with an internal promoter, in which the 3' LTR

could easily be replaced. The structure o f O V A - D is summarized in Fig. 1 (a). In this construct the two LTRs are identical. The U3 promoter is from the SR-D strain o f RSV. The N e o R coding sequence is fused to the first residues of the gag gene, and followed by a mutated version o f the E. coli lacZ gene driven by the SV40 (a)

RSV-DLTR

NeoR

SV40

nlslacZ

RSV-DLTR

6 kb LTR promotedtranscript(NeoRmessengerRNA) 4 kb SV40promotedtranscript(nlslacZmessengerRNA) Probe (b)

Isolde

CEF

kb

kb

-

-

7-4

6F --5.3

4V Fig. 1. Structure of OVA-D: an RSV-derived vector with a fnnctional internal promoter. (a) OVA-D has been designed to produce two types of RNA: a 6 kb transcript initiated at the 5' LTR which can encode the Neo R product or be encapsidated into virus particles, and a 4 kb transcript initiated at the internal SV40 virus promoter which encodes a mutated version of E. coli fl-galactosidase targeted to the cell nucleus. (b) Northern blotting of cellular RNA extracted from Isolde cells stably transfected with OVA-D and CEFs infected with OVA-D. Size markers are indicated on the right. The blot was hybridized to a lacZ antisense RNA probe obtained after in vitro transcription of the C l a ~ B a m H I fragment of OVA-D cloned into pBluescript. Brackets indicate the areas cut out for scintillation counting. After correction for background hybridization, we calculate that 4 kb transcripts represent 30 % of the hybridized RNA.

Self-inactivating avian retrovirus vectors Table 1. Titrations of supernatants harvested from transfected Isolde cells Construct

3' L T R

lacZ titre

Neo N titre

3' L T R structure T A T A box C C A A T box, Enhancer element

SI index*

OVA-D

RSV-D

1"0 x 104

1"0 x 104

1

OVA-R

RAV(1) deleted

1"5 x 102

1"0 x 101

15

OVA-ZZ

RAV(0) tandem

1.0 x 102

1.0 x 102

1

OVA-Zdel

RAV(0) deleted

2.5 x 102

1.0 x 102

2.5

OVA-Zcrip

RAV(0) deleted twice

1-0 x 102

5.0 x 101

2

(a) OVA-D I

* Self-inactivation index: lacZ titre/Neo R titre.

promoter (SVnlslacZ gene). This construct is expected to produce two R N A species: a 6"0 kb transcript initiated at the 5' L T R and covering the entire OVA-D genome, coding for the Neo ~ gene product, and a 4'0 kb R N A initiated at the internal SV40 promoter and coding for fl-galactosidase targeted to the cell nucleus. OVA-D was transfected into Isolde helper cells which were subsequently subjected to G418 selection. The recombinant virus was then harvested from the supernatant of pooled resistant cells and titrated on both CEF and QT6 cells either by selecting with G418 (giving what will be referred to as the Neo R titre) or by using an in situ assay for fl-galactosidase activity (giving the lacZ titre). We obtained an average titre of 104 G418 r.f.u./ml and 104 lac + f.u./ml. These titres are close to that obtained in Isolde cells with NLD, a similar construct which does not have an internal promoter (Cosset et al., 1991). To confirm that avian leukosis virus-derived vectors can accommodate a functional internal promoter, Northern blots were run with RNAs extracted from the transfected Isolde cells and from CEFs infected with OVA-D and selected with G418 (Fig. 1 b). The sizes of the two R N A species detected with a lacZ probe were consistent with the expected pattern of expression. Pieces of the hybridized nylon membrane were then cut out for an approximate R N A quantification. About 30 % of the transcripts appeared to initiate at the internal SV40 promoter. This is consistent with transient expression experiments showing that in QT6 and CEF cells the RSV L T R promoter is more active than the SV40 promoter (data not shown). It appears therefore that avian retrovirus vectors, like murine retrovirus vectors (Cone et al., 1987; Morgenstern & Land, 1990), can accommodate an internal transcription promoter.

Construction of vectors with various Y non-coding sequences To study the possibility of deriving self-inactivating 4

41

-

LTR (b) OVA-R

Neo R SV40 L m k F- ~

nlslacZ

RSV L T R

LTR (c) OVA-ZZ

Neo g SV40 [

ik •

nlslacZ

RAV(1) L T R (deleted)

LTR Neo R SV40.....~ (d) OVA-Zdel [ •

nlslacZ

RAV(0) L T R (tandem) -

_

+

LTR Neo R S V 4 0 _ . . ~ (e) OVA-Zcrip I •

nlslacZ

LTR

nlslacZ RAV(0) L T R (deleted twice)

Neo R SV40

RAV(0) L T R (deleted)

Fig. 2. Structure of the OVA-derived retrovirus vectors. See Methods for the details of constructions. The vectors differ only in their 3" noncoding sequences. White triangles indicate the position of the deletions in the LTR. Angled arrows indicate the position of the SV40 transcription promoter. The presence or absence of enhancer, C C A A T and T A T A elements in the 3' L T R is indicated. As discussed in the text, the absence of an enhancer element in the RAV(0) L T R is questionable.

vectors from OVA-D we designed a series of vectors differing from our first construct only in their Y end (Fig. 2). These vectors were expected to behave similarly when transfected into Isolde cells. After one round o f replication, however, the U3 sequences of both LTRs are expected to derive from the U3 sequence originally located into the Y LTR. Therefore, if the transcription promoter located in this region is inactive, the Neo g expression should be altered in the infected cells. As it is driven by the internal SV40 promoter, the nlslacZ expression should not be affected. Thus, for each vector, the lacZ titre provides us with an estimate of the gene transfer efficiency, and we can define the ratio between the lacZ titre and the Neo ~ titre as a 'self-inactivation index '. Four putative self-inactivating vectors were constructed (Fig. 2). In OVA-R the 3' L T R derives from the RAV(1) virus. A large deletion covers the C C A A T and T A T A boxes but leaves the enhancer element intact. OVA-ZZ possesses an intact RAV(0) L T R (for convenience we kept the L T R tandem repeat present in the original clone). The RAV(0) L T R is present in an endogenous provirus giving rise to relatively high-titre VIR 74

F. Flamant and others

42

Table 2. Clonal variations of virus titre after 0 VA-Zdel

and O VA-Zcrip transfection Clone

lacZ titre

Neo~ titre

SI index

delA delB delC cripA

1.5 x 104 1.6 x 10~ 1.2 x 10~ 2-0 x 10~

2.0 x 103 2.5 x 10z < 20 < 10

7.3 6.4 > 60 > 20

virus (Hughes, 1988) and has been found to possess promoter, terminator, but no enhancer activity (Cullen et al., 1983; H e r m a n & Coffin, 1986). Therefore we expected only a modest inactivation index from this construct, but a good gene transfer efficiency. To improve the self-inactivation index we constructed OVA-Zdel by deleting the RAV(0) L T R from nucleotide - 9 8 to nucleotide - 5 9 (from the cap site), removing the C C A A T box. In OVA-Zcrip the T A T A box was also destroyed from the deleted RAV(0) L T R by site-directed mutagenesis.

be activated when moved into an RSV context (Conklin, 1991). Deleting the 3' C C A A T box in OVA-Zdel resulted in an unexpected improvement in gene transfer efficiency and to a partial self-inactivation. OVA-Zcrip gave lower titres than OVA-Zdel. We isolated 27 clones of helper cells transfected with OVA-Zdel or OVA-Zcrip and titrated their supernatants for nlslacZ and Neo ~ expression. Thirteen of these clones did not give any infectious virus ( < 20 lac + f.u. or G418 r.f.u./ml). This high rate of non-producer clones has been observed with other vectors and might be due to rearrangement occurring in the D N A molecules that are stably transfected into Isolde cells (Cosset et al., 1991). The results obtained with the four clones that gave the highest lacZ titre are reported in Table 2. The selfinactivation index varied widely from one clone to another and in some cases exceeded 60. Therefore the two parameters are not only dependent on the type of vector that is used but also on the producer cell clone.

3" Sequences influence internal gene transient expression Recombinant virus titre and self-inactivation is strongly influenced by 3" non-coding sequences Table 1 summarizes the various titration assays performed with the supernatants obtained from pools of Isolde cells transfected with OVA-D, OVA-R, OVA-ZZ, OVA-Zdel or OVA-Zcrip and further selected with G418 to ensure stable expression of the vector D N A . The gene transfer efficiency was drastically reduced with OVA-R. However, the large deletion that covers both the C C A A T and T A T A boxes led to a partial self-inactivation. More surprisingly, O V A - Z Z which carries a natural 3' L T R had low efficiency. The lack of self-inactivation could be explained by the recently discovered cryptic enhancer element within the RAV(0) U3 sequence, which seems to

To gain further insight into the factors that determine virus production in helper cells we transfected Isolde or QT6 cells with the OVA plasmids and assayed 48 h later for SVnlslacZ transient expression by in situ staining (Table 3). In these conditions OVA-D, O V A - R and O V A - Z Z gave similar results. The deletions within U3 in OVA-Zdel and OVA-Zcrip strongly reduced the SVnlslacZ transient expression. These results support the assumption that the modest virus titre obtained with OVA-Zdel is linked to a defect in SVnlslacZ expression and that the deleted part of the RAV(0) U3 element carries an important signal for viral gene expression. By contrast, the low virus titre obtained with O V A - Z Z does not correlate with the transient nlslacZ expression, indicating that a post-transcriptional step of virus

Table 3. Transient assay for nlslacZ expression Plasmid

3' LTR

OVA-D OVA-R

RSV-D RAV(I) deleted RAV(0) RAV(0) deleted RAV(0) deleted poly(A) MoMuLV:~

OVA-ZZ OVA-Zdel OVA-Zcrip pA0nlslacZ

Calcium phosphate precipitation PolybreneElectroporation Average* 100 NDt"

100 ND

100 75

100 75

100 7

85 14

ND 11

90 10

NO

ND

6

6

100

90

100

95

* ' Average' : ratio to the number of lac+ cells obtained with OVA-D. ? ND, Not determined. :~ MoMuLV, Moloney murine leukaemia virus.

Self-inactivating avian retrovirus vectors

43

OVA-D and OVA-ZZ and lower than that obtained with OVA-Zdel.

50

4O

3' Non-coding sequences influence the efficiency of RNA packaging

o 30 "6

20 10 0 OVA-D

OVA-ZZ

OVA-Zdel

Fig. 3. Clonal variation of nlslacZ expression in infected cells. Clonal variation in the production of OVA-Zdel virus by helper cells. CEFs were infected, selected with G418 and subsequently stained for flgalactosidase activity. The percentage of lac + cells within each resistant clone was then determined: II, 0 % ; Yl, 1 to 5 0 % ; NI, 50 to 100%. Fourteen clones were stained for OVA-D, 16 for O V A - Z Z and 46 for OVA-Zdel (clone delA on Table 2). The distribution was significantly different for OVA-Zdel (P < 5 %).

production is also impaired when using the RAV(0) 3' sequence.

Clonal variation of SVnlslacZ gene expression in infected cells Clones of infected cells resistant to G418 (G418 ~ clones) were stained in situ for fl-galactosidase activity. Although it was performed on a limited number of clones, this analysis led to two interesting observations (Fig. 3). Some G418 R clones contained no lac + cells. This could result from a mutation occurring in the SVnlslacZ gene. However, as G418a/lac helper cells can produce lac + virus (like clone delB in Table 2), it is more likely that G418a/lac - cells actually express SVnlslacZ, but below the in situ detection threshold level. Therefore this assay probably leads to an underestimate of the actual gene transfer efficiency and self-inactivation index. As cells within a given G418 R clone stain with approximately the same intensity, it is likely that chromosomal flanking sequences can greatly influence the SVnlslacZ gene expression. The relative ratio of G418~/lac - to G418~/lac + clones is dependent on the vector used to infect the cells. As expected from transient assays, this ratio is similar for

A low virus titre can result either from a reduced production of virus particles carrying the vector RNA by the helper cells or from a low level of viral gene expression in infected cells. To identify the limiting step in the present case, we analysed retrovirus R N A in helper cells stably transfected by OVA-D, OVA-ZZ and OVA-Zdel (clone delA), and in the virus particles they produce (Table 4). Slot blot analysis showed that the lacZ titre is related to the amount of retrovirus R N A packaged into the virus particles but not to the steadystate level of vector R N A present in the cells. In fact, although OVA-D gave a much higher titre, its RNA steady-state level was not higher within producer cells selected with G418. Quantitative analysis revealed that OVA-ZZ and OVA-Zdel transcripts initiated at the 5' LTR are under-represented in the virus particles. R N A packaging appears therefore to be a limiting step in virus production for OVA-ZZ and OVA-Zdel. Whether this limited reduction in packaging efficiency can entail a significant drop in virus titre remains to be determined.

Expression of Neo R gene in cells infected by 0 VA-Zdel results from rearrangements of the provirus structure Several mechanisms can be proposed to explain how cells infected with our self-inactivating vectors sometimes express the Neo ~ gene. The mutation introduced in the 3' LTR might not fully inactivate the viral transcription promoter. After the transfer of the mutated promoter to the 5' LTR in infected cells, a low level of 5' LTRmediated transcription might occur and sometimes be enhanced by flanking cellular sequences. Alternatively, readthrough transcription from a flanking cellular promoter might be sufficient to ensure Neo R gene expression. Finally our vectors might recombine at high frequency in helper cells or during their reverse transcription and recover a fully active 5' LTR promoter. To distinguish between these possibilities, we compared the proviral structure of OVA-Zdel and OVA-D in pools of infected cells expressing the Neo R gene. The

Table 4. Quantification of RNA packaging Construct

6 kb R N A in cells (%)*

6 kb R N A in virions

Ratio (virion/cell)

Relative ratio (%)*

OVA-D OVA-ZZ OVA-Zdel

100 50 110

9.5 1.9 3.1

9.5 3.8 2.7

100 40 30

* O V A - D was used as a 100 % reference for cellular R N A and packaging ratio. 4-2

44

F. F l a m a n t a n d others

(b)

(a) Acc[

EcoRI

OVA-Zdel in transfeeted helper cells

~

Z

_

OVA-Zdel OVA-D OVA-Zdel OVA-Zdel OVA-D OVA-Zdel infected infected transfected infected infected transfected

I

AccI 2.8 kb

~_

....

kb

]

OVA-Zdel in infected ceils

::~:,:!! ~ ~ ! i

OVA-D in transfected helper cells

2.8 AccI 2.8 kb

~lr

OVA-D after integration in infected cells

2-0

Fig. 4. Structure of the OVA-Zdel provirus 5' LTR in infected cells expressing the NeoR gene. (a) DNA was extracted from pools of cells infected with OVA-Zdel or OVA-D and subjected to G418 selection, and from Isolde cells transfected with OVA-Zdel. After restriction with either AecI or EcoRI and Southern blotting, hybridization was performed with a RNA probe specific for Neor~ sequences. (b) Predicted structure of the integrated proviruses. In infected cells, OVA-D should yield a 2.2 kb EcoRI fragment. The size of the AccI fragment depends on flanking cellular sequences. Cells transfected with OVA-Zdel should yield the same 2.2 kb EcoRI fragment and a 2-5 kb AccI fragment (from the plasmid polylinker to the 3" end of the Neor~ gene). After one round of replication, however, an AccI site should replace the EcoRI site in the 5' LTR. Therefore one expects in infected cells a 2-2 kb AccI fragment and an EeoRI fragment, the size of which is determined by flanking cellular sequences. As a large number of G418-resistant clones were pooled, no EeoRI band should be seen on the autoradiogram. The results rather suggest that, in OVA-Zdel proviruses that express NeoR, the RSV sequences are maintained in the 5' LTR.

Southern blot analysis shown in Fig. 4 reveals the presence o f an E c o R I site in the 5' L T R o f the integrated proviruses and the absence o f an A c c I site. These features strongly suggest that the U3 sequence o f the O V A - Z d e l proviruses that express N e o g does n o t derive f r o m RAV(0) but derives f r o m RSV-D. In fact a N o r t h e r n blot experiment failed to distinguish between the steadystate levels o f b o t h L T R and SV40 transcripts in G418resistant cells infected by either O V A - Z d e l or O V A - D . The fact that the largest messenger is o f definite size also argues against a possible r e a d t h r o u g h transcription f r o m cellular flanking sequences (data n o t shown). Thus it is very likely that O V A - Z d e l can recombine at a high frequency to keep a 5' R S V p r o m o t e r after reverse transcription. The fact that the self-inactivation index is subject to clonal variations a m o n g p r o d u c e r cell clones (see above) suggests that these recombination events are favoured by the highly unpredictable m o d e o f D N A integration that is characteristic o f D N A transfection. Using electroporation, as done in a similar situation (Soriano et al., 1991), did not provide any significant improvement.

Discussion The host-range restriction o f A L V vectors offers a safe alternative to SNV-derived vectors which can infect h u m a n cells ( K o o et aI., 1991). We have constructed two

types o f vector f r o m avian leukosis viruses that could be useful for gene transfer experiments. O V A - D is a highly efficient vector with an SV40 internal promoter. O u r N o r t h e r n blot experiments are consistent with a p r o p e r functioning o f the internal promoter. However, as we did not precisely m a p the transcription initiation o f the l a c Z m R N A , we c a n n o t completely rule out the possibility that l a c Z expression results f r o m an unpredicted splicing o f the L T R - d r i v e n transcript. T w o observations m a d e with O V A - D - r e l a t e d vectors argue against this last hypothesis. First, when self-inactivation is achieved with O V A - Z d e l the l a c Z expression occurs when the L T R - d r i v e n N e o R expression is abolished. Second, when we replaced the SV40 p r o m o t e r by the h u m a n adenovirus E2 promoter, the l a c Z expression became highly inducible by E l a , a transactivator specific for the E2 p r o m o t e r (data not shown). O V A - Z d e l is a vector with a crippled 3" L T R and which has the ability to self-inactivate after replication. The r a n d o m process o f integration o f the vector genome in the transfected helper cells seems however to enable a high rate o f r e c o m b i n a t i o n o f the viral genome, impairing the self-inactivation process. A l t h o u g h we did not identify the recombination mechanism precisely, the fact that the recombination frequency depends on the p r o d u c e r cell clone used leads us to suppose that it is linked to the structure o f the integrated D N A in the transfected cells. F o r example, molecules with a rescued

Self-inactivating avian retrovirus vectors 5' RSV LTR might appear after readthrough transcription across large vector DNA concatemers. These large RNA molecules would then lose the mutated U3 sequence after aberrant splicing or a jump of the reverse transcriptase (Olsen et aI., 1990). This problem can be circumvented by clonal screening of producer helper cells. During our experiments we also found that 3' noncoding sequences can strongly influence not only the selfinactivation ability of retrovirus vectors but also the efficiency of retrovirus-mediated gene transfer. As the U3 deletions that we tested dramatically affected the virus titre, we are prompted to consider that LTR sequences play a major part in this phenomenon. It is known however that the non-coding sequence located upstream of the 3' LTR can also be involved (Sorge et al., 1983; Flanagan et al., 1989). It is difficult to unravel the influence of 3' non-coding sequences on RNA transcription, translation, packaging, and reverse transcription. Previous and present data suggest that several of these parameters can be affected by a 3' mutation. Transient expression experiments have shown that a 3' LTR enhancer element can partially activate the 5' LTR transcription promoter (Norton & Coffin, 1987). If an additive effect of enhancer elements exists, it could explain why OVA-ZZ, OVA-Zdel and OVA-Zcrip yield less IaeZ-expressing virus than OVA-D. However, OVAR does not yield high titres despite the presence of a 3' enhancer. Moreover, Northern blot analysis failed to reveal any influence of 3' sequences on 5' LTR-mediated transcription and led us to suspect a translation defect. Therefore, in the present situation, the 3' LTR enhancer appears to have little or no effect on viral transcription. Recent experiments failed to confirm that a deletion in the downstream LTR can impair the 3' end processing of SNV RNA (Iwasaki & Temin, 1990). Furthermore, it has been shown that RAV(0) LTR sequences are functional in 3' end processing (Herman & Coffin, 1986). Even if readthrough transcription were more frequent with RAV(0)-derived vectors, it would not be expected to have any significant influence on avian retrovirus replication (Swain & Coffin, 1989). Our data show that OVA-ZZ and OVA-Zdel RNAs are slightly under-represented in virus particles, and the same is probably true for OVA-Zcrip. Therefore RNA packaging could be a limiting step in the replication of these vectors. Based on nucleotide sequence analysis, it has been proposed that the 5' leader sequence and the U3 sequence cooperate to build up an RNA secondary structure flavouring encapsidation (Darlix, 1986). As predicted by this model, a deletion in the U5 sequence of the Moloney murine leukaemia virus has been shown to affect RNA packaging (Murphy & Goff, 1989). In OVAZZ and OVA-Zdel the RAV(1) leader sequence would

45

not be able to cooperate with RAV(0) U3 sequences for RNA encapsidation. Indeed RAV(0) and RAV(1) leader and U3 sequences are different in the regions supposed to interact (Bizub et al., 1984; Katz et al., 1986). It seems therefore that the limited lacZ titre obtained with OVA-Zdel might be due to the combination of minor defects in SVnlslacZ gene expression and in RNA packaging. However these considerations cannot explain why the lacZ titre obtained with OVA-ZZ is lower than that obtained with OVA-Zdel. It is likely therefore that another parameter is affected by the U3 deletion. In conclusion, provided that a suitable producer clone is selected, OVA-Zdet appears to be a useful vector that should increase the safety of gene transfer experiments. After replacement of the SV40 promoter by any other promoter, it could also be used to derive new vectors for promoter regulation studies, providing stable integration within the cellular genome in the absence of surrounding viral enhancers.

References AUBERT,D , GARCIA,M., BENCHAIBI,M., PONCET,D., CHEBLOUNE,Y., VERDIER, G., NIGON, V., SAMARUT, J. & MURA, C.V. (1991). Inhibition of primary avian fibroblasts through expression of histone H5 depends on the degree of phosphorylation of the protein. Journal of Cellular Biology 113, 497-506. BENCHAIBI,M., MALLET,F., THORAVAL,P., SAVATIER,P., XIAO, J. H., VERDIER,G. & NIGON,V. M. (1989). Avian retroviral vectors derived from avian defective leukemia virus: role of the transcriptional context of the inserted gene on efficiency of the vectors. Virology 169, 15-26. BIZUB, D., KATZ, R. A. & SKALKA,A. M. (1984). Nucleotide sequence of non-coding regions in Rous-associated virus-2: comparisons delineate conserved regions important in replication and oncogenesis. Journal of Virology 49, 557-565. BONNEROT,C., ROCANCOURT,D., BRIAND,P., GRIMBER,G. & NICOLAS, J. F. (1987). A fl-galactosidase hybrid protein targeted to nuclei as a marker for developmental studies. Proceedings of the National Academy of Sciences, U.S.A. 84, 6795 6799. CONE, R. C , WEBER-BENAROUS,A., BAORTA,D. & MULLIGAN,R. C. (1987). Regulated expression of a complete human fl-giobin gene encoded by a transmissible retrovirus vector. Molecular and Cellular Biology 7, 887-897. CONKLIN, K.F. (1991). Activation of an endogenous retrovirus enhancer by insertion into a heterologous context. Journal of Virology 65, 25252532. COSSET, F. L., LEGRAS,C., CHEBLOUNE,Y., SAVATIER,P., THORAVAL, P., THOMAS,J. L., SAMARUT,J., NIGON, V. & VERDIER,G. (1990). A new avian leukosis virus (ALV)-based packaging cell line that uses two separate transcomplemenfing helper genomes. Journal of Virology 64, t070-1078. COSSET,F. L., LEGRAS,C., THOMAS,J. L., MOLINA,R. M., CHEBLOUNE, Y., FAURE,C., NIGON,V. M. & VERDIER,G. (1991). Improvement of avian leukosis virus (ALV)-based vectors by using different cis-acting sequences from ALV. Journal of Virology 65, 338~3394. CULLEN, B. R., SKALKA,A. M. & Ju, G. (1983). Endogenous avian retroviruses contain promoter and leader sequences. Proceedings of the National Academy of Sciences, U.S.A. 80, 294(~2950. DARLIX, J. L. (1986). Circularizafion of retroviral genomic RNA and the control of RNA translation, packaging and reverse transcription. Biochimie 68, 941-949. DOUGHERTY, J. P. & TEMIN, H. M. (1987). A promoterless retroviral vector indicates that there are sequences in U3 required for 3' RNA

46

F. Flamant and others

processing. Proceedings of the National Academy of Sciences, U.S.A. 84, 1197-1201. EGLITIS, M.A. & ANDERSON, W.F. (1988). Retroviral vectors for introduction of genes into mammalian cells. Biotechniques 6, 608-614. EMERMAN, M. & TEMIN, H.M. (1984). Genes with promoters in retrovirus vectors can be independently suppressed by an epigenetic mechanism. Cell 39, 459-467. EMERMAN, M. & TEMIN, H. M. (1986). Quantitative analysis of gene suppression in integrated retrovirus vectors. Molecular and Cellular Biology 6, 792-800. FLANAGAN, J. R., KRIEG, A. M., MAX, E. E. & KHAN, A. S. (1989). Negative control region at the 5' end of murine leukemia virus long terminal repeats. Molecular and Cellular Biology 9, 739-746. GORMAN, C.M., MERLINO, G.T., WILLINGHAM, I., PASTAN, J. & HOWARD, B. H. (1982). The Rous sarcoma virus long terminal repeat is a strong promoter when introduced in a variety of eukaryotic cells by DNA mediated transfection. Proceedings of the National Academy of Sciences, U.S.A. 79, 6777-6781. HERMAN, S. A. & COFFIN, J. M. (1986). Differential transcription from the long terminal repeats of integrated avian leukosis virus. Journal of Virology 60, 497-505. HUGHES, S.H. (1988). Retroviral vectors and adaptors. In Viral Vectors. Edited by Y. Gluzman & S. H. Hughes. New York: Cold Spring Harbor Laboratory. IWASAKI,K. & TEMIN, H. M. (1990). The U3 region is not necessary for 3' end formation of spleen necrosis virus RNA. Journal of Virology 64, 6329-6334. KATZ, R. A., CULLEN, B. R., MALAVARCA,R. & SKALKA,A. M. (1986). Role of the avian retrovirus mRNA leader in expression: evidence for novel translational control. Molecular and Cellular Biology 6, 372 379. KAWAI, S. ~; NISHIZAWA,M. (1984). New procedure for DNA transfection with polycation and dimethyl sulfoxide. Molecular and Cellular Biology 4, 1172-1174. Koo, H. M., BROWN, A. M. C., RON, Y. &; DOUGHERTY, J. P. (1991). Spleen necrosis virus, an avian retrovirus, can infect primate cells. Journal of Virology 65, 4769-4776. MANIATIS, T., FRITSCH, E.F. &; SAMBROOK, .l. (1982). Molecular Cloning: A Laboratory Manual. New York: Cold Spring Harbor Laboratory. MORGENSTERN, P. J. & LAND, H. (1990). Advanced mammalian gene transfer: high titre retroviral vectors with multiple drug selection markers and a complementary helper-flee packaging cell line. Nucleic Acids Research 18, 3587-3596. MOSCO¥1CI, C., MOSCOVICI, G., JIMENEZ, H., LAI, M. M. C., HAYMAN, M.J. & VOGT, P.K. (1977). Continuous tissue culture cell lines

derived from chemically induced tumors of Japanese quail. Cell 11, 95-103. MURPHY, J. E. & GORE, S. (1989). Construction and analysis of deletion mutations in the U5 region of Moloney murine leukemia virus: effects on RNA packaging and reverse transcription. Journal of Virology 63, 319-327. NORTON, P.A. & COFFIN, J.M. (1987). Characterization of Rous sarcoma virus sequences essential for viral gene expression. Journal of Virology 61, 1171-1179. OLSEN, J.C., BOVA-HILL, C., GRANDGENETT, D.P., Qur~N, T.P., MANFREDI, J.P. t~; SWANSTROM, R. (1990). Rearrangements in unintegrated retroviral DNA are complex and are the result of multiple genetic determinants. Journal of Virology 64, 5475-5484. OLSON, P., TEMIN, H.M. & DORNBURG, R. (1992). Unusually high frequency of reconstruction of long terminal repeats in U3-minus retrovirus vectors by DNA recombination or gene conversion. Journal of Virology 66, 1336-1343. PAYNE, G. S., COURTNEIDGE, S. A., CRITTENDEN, L. B., FADLY, A. M., BISHOP, J. M. & VARMUS, H. M. (1981). Analysis of avian leukosis DNA and RNA in bursal tumors: viral gene expression is not required for the maintenance of tumor state. Cell 23, 311-321. SANES,J. R., RUBENSTEIN, J. R. L. & NICOLAS, J. F. (1986). Use of a recombinant retrovirus to study post-implantation cell lineage in mouse embryos. EMBO Journal 5, 3133-3142. SAVATIER, P., BAGNIS, C., THORAVAL, P., PONCET, D., BELAKEBI,M., MALLET, F., LEGRAS, C., COSSET, F. L., THOMAS, J. L., CHEBLOUNE, Y., FAURE, C., VERDIER, G., SAMARUT, J. • NIGON, V. (1989). Generation of a helper cell line for packaging avian leukosis virus based vectors. Journal of Virology 63, 513-522. SORGE, J., RICCI, W. 8,: HUGHES, S.H. (1983). Cis-acting RNA packaging in the 115-nucleotide direct repeat of Rous sarcoma virus. Journal of Virology 48, 667-675. SORIANO, P., FRIEDRICH, G. & LAWINGER, P. (1991). Promoter interactions in retrovirus vectors introduced into fibroblasts and embryonic stem cells. Journal of Virology 65, 2314-2319. SWAIN,A. & COFFIN, J. M. (1989). Polyadenylation at correct sites in genome RNA is not required for retrovirus replication or genome encapsidation. Journal of Virology 63, 3301-3306. YEE, J. L., MOORES, J. C., JOLLY, D. J., WOLFF, J. A., RESPESS,J. G. & FRIEDMANN, T. (1987). Gene expression from transcriptionally disabled retroviral vectors. Proceedings of the National Academy of Sciences, U.S.A. 84, 5197-5201. Yu, S. L., VON RUDEN, T., KANTOFF,P. W., GARBER,C., SEIBERG,M., RUTHER, U., ANDERSON, W. F., WAGNER,E. F. & GILBOA, E. (1986). Self-inactivating retrovirat vectors designed for transfer of whole genes into mammalian cells. Proceedings of the National Academy of Sciences, U.S.A. 83, 3194-3198.

(Received 8 June 1992; Accepted 8 September 1992)