Vol. 65, No. 12

JOURNAL OF VIROLOGY, Dec. 1991, p. 6942-6952 0022-538X/91/126942-11$02.00/0 Copyright © 1991, American Society for Microbiology

Circularization of Human Immunodeficiency Virus Type 1 DNA In Vitro CHRIS M. FARNET AND WILLIAM A. HASELTINE* Division of Human Retrovirology, Dana-Farber Cancer Institute, and Department of Pathology and Committee on Virology, Harvard Medical School, 44 Binney Street, Boston, Massachusetts 02115 Received 13 June 1991/Accepted 16 September 1991

Linear viral DNA present in cytoplasmic extracts of cells newly infected with human immunodeficiency virus can be induced to form 1-LTR and 2-LTR circles by incubation of the extracts in the presence of added nucleoside triphosphates. No circular DNA forms are detected when extracts are incubated in the absence of added nucleoside triphosphates. Restriction enzyme analysis and polymerase chain reaction analysis with selected primers, as well as DNA sequence analysis of the polymerase chain reaction products, show that most of the 2-LTR circles are the result of autointegration reactions, while 1-LTR circles result from recombination between the long terminal repeats on the linear viral DNA. In addition, a small amount of simple 2-LTR circles, formed by end-to-end joining of the linear viral DNA, is formed in vitro. Integration of the linear viral DNA into heterologous DNA competes effectively with the formation of 2-LTR circles by autointegration. However, concentrations of target DNA which completely block autointegration have no effect on the formation of 1-LTR circles or simple 2-LTR circles. Factors present in extracts of uninfected cells can mediate the formation of 1-LTR circles and simple 2-LTR circles from purified deproteinated linear viral DNA, indicating that viral proteins are not necessary for the formation of these two types of circular viral DNA. These experiments demonstrate that all the transformations of linear viral DNA which occur in the nuclei of cells infected with human immunodeficiency virus type 1 can be reproduced in vitro. type 1

Cells productively infected with retroviruses harbor a variety of DNA molecules derived from the infecting viral RNA genome. In the cytoplasm of the cell, reverse transcription of the viral RNA generates a linear, doublestranded DNA molecule containing the viral genes bounded by directly repeated sequences termed long terminal repeats (LTRs) (31). The nucleus of the infected cell contains several forms of viral DNA in addition to the linear DNA synthesized in the cytoplasm (Fig. 1). Viral DNA integrated into the host genome is identical in structure to the unintegrated linear molecule, except for the absence of 2 bp from each LTR terminus at the sites of joining to host DNA (30). At least two forms of circular viral DNA are also found in the nucleus. The most abundant form contains a single copy of the LTR, while a smaller number of circles contain two LTRs (24, 25, 34). Recent studies in a number of retroviral systems have provided convincing evidence that the linear unintegrated form of viral DNA is the direct precursor to the integrated provirus (4, 12). The development of in vitro integration systems and the analysis of in vitro reaction intermediates have provided detailed information on the mechanism of provirus formation. Comparatively little is known about the biochemical events involved in the circularization of viral DNA. While it has been demonstrated that the linear viral DNA synthesized in the cytoplasm is the precursor to the circular DNA forms found in the nucleus (25), the mechanism of circularization remains unknown. Circles containing one LTR have been proposed to arise either from homologous recombination between the LTRs present on the linear viral DNA molecule (13, 24) or from a circular intermediate in the formation of the linear molecule during reverse transcription (7, 14). Circles with two tandem LTRs, on the *

Corresponding author. 6942

other hand, are presumed to arise by the direct ligation of the ends of the linear DNA molecule (15, 29). Yet another class of circular DNA molecules containing two LTRs has been identified that appears to arise from intramolecular integration (autointegration) events, presumably mediated by the viral integrase (17, 26, 27). Time course analysis of the appearance of viral DNA forms in the nucleus following a single round of infection indicates that the linear viral DNA molecules undergo rapid transformation soon after entry into the nucleus (9, 16, 25). The concentration of linear DNA in the nucleus remains relatively constant and low, while the unintegrated circular forms and integrated proviruses accumulate with time (9, 11, 16, 25). These observations suggest that once the linear viral DNA enters the nucleus, it either stably integrates into the host DNA or forms circular molecules which no longer have the capacity to integrate. Factors which influence viral integration and circle formation are difficult to study in infected cells. Previously, it was demonstrated that linear viral DNA present in cytoplasmic extracts of cells newly infected with human immunodeficiency virus type 1 (HIV-1) integrates with high efficiency into heterologous target DNA molecules in vitro (9). Here, it is demonstrated that the viral DNA in these extracts can also be induced to form both 1-LTR and 2-LTR circles. Analyses of the in vitro reaction conditions which favor viral DNA integration or circularization suggest that integration and circle formation are two independent and competing fates for the newly formed linear viral DNA. MATERIALS AND METHODS

Preparation of cell extracts. The chronically HIV-1 infected Molt IIIB human T-cell line (HTLV-IIIB virus strain) was the source of virus for cell-free infections. Culture supernatant containing a high concentration of virus was

VOL. 65, 1991

CIRCULARIZATION OF HIV-1 DNA IN VITRO

6943

linear vira DNA

1 a .

v

PMI

A 8M12*

f xSa I, Pat I

I

U

IR GUl"i

integrated provirus I U3 IRLI d amI t3IR

A

1415bp

2.

PM'

Im I U3 IR hsl 3mt4p SAI

3.

4. simple 2 - LTR circles MI

AZ

xSa

PO

U51 R US I

JIMIJRI IU5IRI US I J

SolI

1-LTRjuncton

5.

A

A

A

47121p

4

U51 R

f xSaJ I, Pst I

;

A

Psti

1, PstI

5IR I

Pati

a

SidI

5393 bp

2-LTR junclon

6.

,&xSaI 1, Pst I heterogeneous LTR-containing

fragments

xSafI, Pet I heterogeneous LTR-containing

fragments FIG. 1. Structures and mechanisms of formation of the viral DNA forms found in an infected cell. Linear viral DNA (structure 1) is synthesized in the cytoplasm of the cell shortly after infection and moves into the nucleus, where several additional types of viral DNA are found. 1-LTR circles (structure 3), simple 2-LTR circles (structure 4), and circular products resulting from autointegration (structures 5 to 8) can be found among the unintegrated viral DNA forms in the nucleus. In addition, some of the linear viral DNA integrates into the host genome to form the provirus (structure 2). Structures 5 and 6 depict the two possible orientations of autointegration events. The joining of opposite DNA strands at the site of integration (structure 5) yields 2-LTR circular molecules having inversions of viral DNA sequences bordered by nontandem LTRs (structure 7). Joining of the same DNA strands at the site of integration (structure 6) results in the formation of two independent (although possibly catenated) circles, each of which contains one LTR (structure 8). The LTR-containing products expected for each type of unintegrated viral DNA following digestion with SalI and PstI are also shown. Note that only the LTR-containing restriction products are shown, for simplicity. Positions of primers used for PCR amplification are indicated as arrows labeled PBS, tat, nef, Rl, and R2. The asterisks in products 7 and 8 denote the new LTR-target junctions formed by the autointegration reaction. Note that only one of the two DNA strands is most likely joined at each integration junction formed in vitro (see references 4 and 12); this likelihood does not affect the subsequent PCR amplification reactions, as only a single continuous DNA strand is needed for amplification across the junction.

6944

FARNET AND HASELTINE

prepared by incubating 300 x 106 Molt IIIB cells in 50 ml of fresh RPMI 1640 medium containing 20% fetal calf serum and 10 ng of phorbol 12-myristate 13-acetate per ml for 20 to 24 h. Cells were removed by centrifugation, and the supernatant, containing 15 to 20 ,ug of p24 viral core protein per ml, was used to resuspend 100 x 106 cells of the human SupTl T-cell line. Cytoplasmic extracts (3 ml) were prepared 4 to 4.5 h postinfection as described previously (9). Cell extracts from uninfected SupTl cells were prepared in the same manner.

Circularization of linear viral DNA was induced by the addition of nucleoside triphosphates (NTPs) to cell extracts. All NTPs, deoxyribonucleoside triphosphates (dNTPs), and

dideoxyribonucleoside triphosphates (ddNTPs) were purchased from Pharmacia LKB Biotechnology, Inc., Piscataway, N.J. Azidothymidine was a gift from Burroughs Wellcome Co., Research Triangle Park, N.C. ATP-yS and all other nucleosides and related compounds were purchased from Sigma Chemical Co., St. Louis, Mo. Fractionation of cell extracts. Viral preintegration complexes were partially purified from cytoplasmic extracts by Sephacryl S400 gel filtration chromatography as described previously (10). Purification of linear viral DNA. Linear viral DNA was

prepared from infected cell extracts by sodium dodecyl sulfate (SDS)-proteinase K treatment and phenol extraction as described previously (9). For the formation of circular DNA in vitro, linear viral DNA prepared from 0.32 ml of extract was resuspended in 10 ,ul of 10 mM Tris HCl (pH 7.4) and added to 0.32 ml of cytoplasmic extract prepared from uninfected SupTl cells. ATP or another nucleotide was added to a final concentration of 1 mM, and the mixture was incubated at 37°C. Viral DNA was purified from the extracts described below. Analysis of viral DNA. Viral DNA was purified from extracts and reaction mixtures by SDS-proteinase K treatment and phenol extraction as previously described (9). Total unintegrated viral DNA was prepared from SupTl cells 24 h postinfection by a modified Hirt extraction (6). Agarose gel electrophoresis and Southern blotting and hybridization analyses of viral DNA were performed as previously described (9). PCR analysis and DNA sequencing. The following oligonucleotides were used for polymerase chain reaction (PCR): PBS, 5'-GTCGCCGCCCCTCGCCTC-3'; tat, 5'-TTCTGAT GAGCTCTTCGTCGC-3'; nef, 5'-GGGGGATCCGAAGAAG AAGGTGGAGAGCGA-3'; Rl, 5'-TGGCTAACTAGGGAAC CCACTGCTTAAGCC-3'; R2, 5'-AGAGCTCCCAGGCTCA GATCTGGTCTAACC-3'. The locations of these primers in the viral genome are indicated in Fig. 1. Reaction mixtures (100 ,ul) contained 10 mM Tris HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl2, 0.001% (wt/vol) gelatin, 200 FM each dATP, dCTP, dGTP, and TTP, 1 ,uM each indicated primer, and 2.5 U of AmpliTaq recombinant Taq DNA polymerase (PerkinElmer Cetus Co., Norwalk, Conn.). For amplification of viral DNA from cytoplasmic extracts, each reaction mixture contained approximately 1/30th the amount of DNA from a 3-ml extract. For amplification of viral DNA from Hirt extractions, each reaction mixture contained DNA from the equivalent of approximately 2 x 106 cells. An initial denaturation step at 98°C was performed for 10 min. Thirty cycles of the PCR were then performed with an annealing temperature of 60°C for 1 min, an extension temperature of 72°C for 5 min, and a denaturation temperature of 95°C for 1 min. Reaction products were visualized after electrophoresis on 1% agarose gels containing ethidium bromide. For DNA as

J. VIROL.

sequence analysis of autointegration events, the smear of DNA products generated by PCR was cut out of the gel and the DNA was purified from gel slices by using Geneclean (Bio 101, Inc., La Jolla, Calif.) and sequenced directly by the methods of Sanger et al. (23), using the following primer, located in the U3 region of the virus: 5'-GCCCTGGTGTG TAGTTCTGC-3'.

RESULTS Circularization of viral DNA in vitro. Cell extracts were prepared from cells 4 to 5 h after infection with HIV-1, using methods previously demonstrated to yield fully functional viral preintegration complexes (9). Viral DNA present in the cytoplasmic extracts migrated as a single species when analyzed by agarose gel electrophoresis and Southern hybridization (Fig. 2A, lane 1). The extracts contained no detectable circular viral DNA, consistent with previous reports that only full-length linear viral DNA is formed in the cytoplasm of cells newly infected with HIV-1 (9, 16). Viral DNA present in the cytoplasmic extracts integrated into target DNA in vitro with very high efficiency (Fig. 2A, lane 3). The observation that circular DNA forms can be generated in vitro was made upon analysis of the reaction products made in the presence of ATP. In these reactions, ATP was added to the cytoplasmic extract in the absence of added target DNA. The data presented in Fig. 2 show that the addition of 1 mM ATP to cytoplasmic extracts resulted in the formation of two viral DNA species that migrated more slowly than the linear form of viral DNA on agarose gels (Fig. 2A, lane 4). Formation of the slowly migrating DNA forms depended on the addition of ATP; viral DNA in extracts incubated at 37°C in the absence of added ATP remained linear (Fig. 2A, lane 2). The electrophoretic mobilities of the two new DNA forms were identical to those observed for the 1- and 2-LTR open circular viral DNA molecules found in Hirt supernatants prepared from acutely infected cells (Fig. 2A, lane 5). Digestion of the reaction products with SalI, which recognizes a single site in viral DNA, generated two novel DNA forms, not seen in SalI digests of linear viral DNA, having electrophoretic mobilities expected of linear DNA molecules containing one or two LTRs (Fig. 2B, lane 6). DNA molecules of the same size were observed following Sall digestion of viral DNA from Hirt supernatants (Fig. 2B, lane 8). These observations confirm the circular nature of the slowly migrating DNA forms generated by incubation of the extracts in the presence of added ATP. Incubation of the cytoplasmic extracts with 1 mM ATP-yS, a nonhydrolyzable analog of ATP, resulted in the formation of only the slower migrating of the two new DNA forms (Fig. 2B, lane 2, and C, lane 4). Digestion of the reaction products formed in the presence of ATP-yS with SalI generated a linear molecule having the same size as full-length linear viral DNA, indicating that the slower migrating of the new DNA forms contained two copies of the viral LTR sequences (Fig. 2B, lane 5). Restriction enzyme analysis of circular viral DNA. The different forms of circular viral DNA can be distinguished by analysis of the LTR-containing fragments generated by restriction endonuclease digestion (diagrammed in Fig. 1). Digestion of circular viral DNA with enzymes that cut at sites bordering the LTRs is predicted to yield two DNA fragments, not found in digests of linear viral DNA, that differ in size by the length of the LTR. The smaller fragment, containing a single copy of the LTR, is derived from the

VOL. 65, 1991

CIRCULARIZATION OF HIV-1 DNA IN VITRO

A.

B. 1 2 3 4 5

C. 1 2 3 4 5 6 7 8

1

IP -23.1

65

-- 4.3

2 3 4 5 6 7 8 9 10 11

23.1 w

-

6945

**

--

9.4

--

6.5

_

,,.gw,b

43

--- 23.1

*434

9.4 65 f

_

~~~~~--.4.3

FIG. 2. Southern hybridization and restriction endonuclease analyses of viral DNA prepared from cytoplasmic extracts 4 h after infection with HIV-1, as described in Materials and Methods. In all cases, an HIV-1 LTR-containing DNA fragment was used as a hybridization probe (9). Therefore, only LTR-containing DNA fragments are detected. (A) Viral DNA prepared from untreated extracts (lane 1), extracts incubated at 37°C for 45 min (lane 2), extracts incubated with 650 ng of linear 4X174 DNA per ml at 37°C for 45 min (lane 3), and extracts incubated with 1 mM ATP at 37°C for 45 min (lane 4). Lane 5, unintegrated viral DNA prepared by a modified Hirt extraction (6) of SupTl cells infected 24 h earlier with HIV-1. In each lane, the band migrating at approximately 9.7 kb represents the linear form of unintegrated viral DNA. In addition, in lane 3, the band migrating at approximately 15 kb represents linear viral DNA integrated into the linear 4X174 DNA target, and in lanes 4 and 5, the bands migrating at approximately 23.1 and 24.5 kb represent the 1-LTR and 2-LTR open circular forms of viral DNA, respectively. (B) Restriction endonuclease analysis of circularization reaction products. Viral DNA was purified from extracts incubated with 1 mM ATP for 15 min (lanes 1 and 4), with 1 mM ATP for 45 min (lanes 3 and 6), or with 1 mM ATP-yS for 45 min (lanes 2 and 5). Lanes: 1 to 3, uncut viral DNA; 4 to 6, viral DNA cut with Sall prior to electrophoresis; 7, linear viral DNA prepared from untreated extracts and digested with SalI prior to electrophoresis; 8, unintegrated viral DNA prepared by Hirt extraction of infected cells, as described in panel A, digested with Sall prior to electrophoresis. (C) Time course and restriction endonuclease analysis of circularization reaction products. Lanes: 1 to 3, viral DNA prepared from extracts treated with 1 mM ATP for 15, 30, and 45 min, respectively; 4, viral DNA prepared from extract treated with 1 mM ATP-yS for 45 min; 5, linear viral DNA prepared from untreated extracts and digested with SalI and PstI prior to electrophoresis; 6 to 8, viral DNA from extracts shown in lanes 1 to 3, digested with SalI and PstI prior to electrophoresis; 9, Sall-PstI digestion of viral DNA prepared from extract treated with 1 mM ATP for 90 min; 10, Sall-PstI digestion of reaction products shown in lane 4; 11, SalI-PstI digestion of unintegrated viral DNA prepared by Hirt extraction of infected cells. In all panels, numbers on right show size in kilobases.

1-LTR circular viral DNA. The larger fragment contains two tandem copies of the LTR and is derived from 2-LTR circles formed by end-to-end joining of the linear viral DNA (hereafter referred to as simple 2-LTR circles). Two types of circular DNA products are predicted to result from the autointegration of viral DNA depending on the orientation of the DNA strand-joining reaction (Fig. 1). Joining of the 3'-hydroxyl terminus of each LTR to the 5'-phosphate of the opposite strand of viral DNA at the site of target cleavage will result in the formation of a circular molecule containing an inversion of viral sequences between two nontandem LTRs (Fig. 1, product 7). These products will be linearized by restriction enzymes that cut viral DNA once, yielding molecules having the same size as the full-length linear form of viral DNA. Digestion of these circular products with enzymes that cut viral DNA more than once is expected to yield LTR-containing DNA fragments of heterogeneous lengths, as the site of integration within the viral DNA should not be constant. Alternatively, the autointegration reaction may result in the joining of the 3'-hydroxyl terminus of each LTR to the 5'-phosphate of the same strand of viral DNA at the site of target cleavage (Fig. 1, product 8). This type of event should yield a pair of circles of various sizes, each of which will be smaller than genome length and contain a single LTR. Digestion of such circles with restriction enzymes that cut viral DNA either singly or multiply is predicted to yield a heterogeneous mixture of LTR-containing DNA fragments. The structures of the circular DNA molecules formed in vitro were analyzed by restriction enzyme digestion and compared with the structures of circular molecules formed in vivo. Digestion of the in vitro circularization reaction products with SaI and PstI, which each cut viral DNA once, generated two novel LTR-containing junction fragments not present in digests of linear viral DNA (Fig. 2C, lanes 5 to 9).

These fragments had the same size as the LTR-containing junction fragments produced by digestion of viral DNA prepared from Hirt supernatants of acutely infected cells (Fig. 2C, lane 11). The autoradiographic intensity of the 1-LTR junction fragment produced by digestion of the in vitro circularization reaction products was identical to the intensity of the uncut 1-LTR circular molecule. In contrast, the intensity of the fragment encoding the 2-LTR junction was reproducibly severalfold lower than that of the uncut 2-LTR circle. Time course analysis demonstrated that the amount of 1-LTR junction fragment formed in the in vitro reaction increased in proportion to the amount of 1-LTR circular DNA formed (Fig. 2C, compare lanes 1 to 3 and 6 to 8). On the other hand, the amount of 2-LTR junction fragment increased at a rate far slower than the rate of increase of the amount of 2-LTR circular DNA. These results indicate that all the 1-LTR circular molecules formed in vitro resulted from the formation of a novel 1-LTR junction, while only a small fraction of the 2-LTR circles formed in vitro could be explained by simple end-to-end joining of the linear viral DNA. In addition, digestion of the circularization reaction products with these enzymes generated higher hybridization backgrounds than digestion of linear viral DNA (Fig. 2C, lane 5), consistent with the presence of a heterogeneous population of circular molecules generated by autointegration. As mentioned previously, circular viral DNA formed by the addition of ATP-yS to cell extracts was exclusively of the 2-LTR type, with no detectable 1-LTR circles produced. Digestion of ATPyS-induced circles with Sall and PstI demonstrated the absence of 1-LTR junction fragments (Fig. 2C, lane 10). In addition, no new 2-LTR junction fragments were detected, demonstrating the absence of simple 2-LTR circles among the reaction products. Instead, a high background of heterogeneous LTR-containing fragments was

6946

*. :4 ,. :Vi.-

FARNET AND HASELTINE

A.

B.

1 2 3 4 5 6 7 8 9 .....

*:~ ;.

A.Wll,i-,i..;-: -ww i;

4" W

iow .00,

. -

''

J. VIROL. 1

1 2 3 4 5 6

0

3

23.1

- 23.1

9.4

9.4

2 3 4 5 6 7 8 9 10 11 12 13 14 -,

-.0 ver, mt v

.,

23.1

* .I9.4 ..~*_t-

FIG. 3. Effect of ATP and KCI concentrations on in vitro circularization reactions. (A) Cytoplasmic extracts were prepared 4 h postinfection, and ATP was added to a final concentration of 0.5 mM (lanes 2 and 6), 1 mM (lanes 3 and 7), 2 mM (lanes 4 and 8), or 5 mM (lanes 5 and 9) before incubation at 37°C for 45 min (lanes 2 to 5) or 90 min (lanes 6 to 9). Lane 1, viral DNA prepared from extract incubated at 37°C for 90 min in the absence of added ATP. (B) Cytoplasmic extracts were adjusted to a final KCI concentration of 100 mM (lane 1), 125 mM (lane 2), 150 mM (lane 3), 175 mM (lane 4), 200 mM (lane 5), or 250 mM (lane 6), and ATP was added to a final concentration of 1 mM before incubating at 37°C for 90 min. In both panels, numbers on right show size in kilobases.

produced, indicating that the 2-LTR circles formed in the presence of the nonhydrolyzable ATP analog were entirely the products of autointegration. Conditions affecting the circularization of viral DNA in vitro. Analysis of the conditions affecting the circularization of viral DNA in vitro indicated that the two circular viral DNA forms were generated by distinct mechanisms. The formation of each circular species had a distinct ATP concentration optimum. Formation of 2-LTR circles was optimal in extracts containing 1 mM ATP and was completely inhibited by a concentration of 5 mM ATP (Fig. 3A). In contrast, 1-LTR circle formation peaked at 2 mM ATP and was considerable even at 5 mM ATP (Fig. 3A). Lower salt concentrations favored the formation of 1-LTR circles (Fig. 3B). At a concentration of 100 mM KCl, 1-LTR circles were formed almost exclusively after incubation with ATP (Fig. 3B, lane 1). The most efficient formation of 1-LTR circles occurred in extracts containing 125 mM KCl (Fig. 3B, lane 2). In contrast, 2-LTR circle formation was most efficient at KCI concentrations between 150 and 175 mM (Fig. 3B, lanes 3 and 4). At the higher salt concentrations, circular reaction products were almost entirely of the 2-LTR form (Fig. 3B, lanes 5 and 6). The formation of both forms of circular viral DNA was dependent on the presence of magnesium. Addition of 5 mM EDTA to extracts at the time of ATP addition abolished detectable circle formation (Fig. 4, lane 14). In addition, formation of both circular forms depended on protein functions present in the extracts, as pretreatment of extracts with proteinase K for 30 min prior to the addition of ATP abolished circle formation (Fig. 4, lane 13). A variety of nucleotides and related compounds were tested for the ability to induce the circularization of viral DNA in vitro. dATP induced both circular forms with an efficiency comparable to that of ATP (Fig. 4, lane 6). All the other NTPs and dNTPs induced the formation of 2-LTR circles but induced the formation of only a small amount of the 1-LTR circular form (Fig. 4, lanes 2 to 5, and data not shown). Similarly, the 2',3'-ddNTPs, ddATP, ddCTP, ddGTP, and ddTTP, induced the formation of 2-LTR circles but were unable to induce detectable amounts of the 1-LTR circular forms (Fig. 4, lane 2, and data not shown). Only background levels of circular DNA forms were observed as

FIG. 4. Testing the ability of other compounds to induce the circularization of viral DNA in vitro. Lanes 1 to 9, cytoplasmic extracts were incubated with added ATP (lane 1), ddATP (lane 2), GTP (lane 3), CTP (lane 4), UTP (lane 5), dATP (lane 6), ADP (lane 7), AMP (lane 8), or adenosine (lane 9). In all cases, compounds were added to a final concentration of 1 mM and incubated at 37°C for 45 min, except for the reaction shown in lane 1, which was incubated for 90 min. Lanes 10 to 12, extracts were incubated for 45 min at 37°C after the addition of 1 mM ATP (lane 10), 10 mM phosphocreatine and 200 ,ug of creatine phosphokinase per ml (lane 11), or 10 mM phosphocreatine (lane 12). Lane 13, proteinase K was added to the cytoplasmic extract and incubated at 37°C for 10 min, and then ATP was added to 1 mM final concentration and incubation continued for 45 min. Lane 14, 5 mM EDTA and 1 mM ATP were added simultaneously to the cytoplasmic extract and then incubated at 37°C for 45 min. Numbers on right show size in kilobases.

products of reactions containing the nucleoside diphosphates ADP and GDP, the nucleoside monophosphate AMP, the nucleosides adenosine, guanosine, cytosine, azidothymidine, and cordycepin, or the free bases adenine, guanine, and uracil (Fig. 4, lanes 7 to 9, and data not shown). Both 1-LTR and 2-LTR circles were formed following the addition of phosphocreatine and creatine phosphokinase to the cytoplasmic extracts, presumably because of the synthesis of ATP from endogenous ADP (Fig. 4, lane 11). PCR analysis of circular viral DNA. PCR analysis of the circularization reaction products also provided evidence for the formation of 2-LTR circles by autointegration. The primers used for detection of autointegration events are indicated in Fig. 1. Both amplification primers used anneal to the sense strand of viral DNA, one in the viral primerbinding site (PBS) region upstream of the gag gene, the other in the tat gene. Since they anneal to the same strand of DNA, these primers will not amplify any sequences on linear or 1-LTR circular viral DNA or 2-LTR circles formed by end-to-end joining. However, these primers will anneal to opposite DNA strands on those circular autointegration products that result from the joining of opposite strands of viral DNA at the site of integration (Fig. 1, product 7). Therefore, 2-LTR circles formed by autointegration will direct the amplification of a population of molecules of heterogeneous length when the PBS and tat primers are used. All the amplified sequences are predicted to have a full LTR sequence at the end defined by the PBS amplification primer (as diagrammed in Fig. SC). PCR amplification of viral DNA purified from extracts that had been incubated with 1 mM ATP using the PBS and tat primers produced a mixture of DNA fragments of heterogeneous lengths (Fig. 5A, lane 1). In contrast, no detectable products were made when viral DNA from extracts incubated at 37°C for 1 h in the absence of added ATP was the substrate for PCR (Fig. SA, lane 2). Digestion of the amplification products with MroI, an enzyme that cuts at a single site in the U3 region of the viral LTR, generated a 425-bp DNA fragment, demonstrating that the heterogeneous population of amplified molecules contained a viral LTR at one end (Fig. SA, lane 3).

CIRCULARIZATION OF HIV-1 DNA IN VITRO

VOL. 65, 1991

D.

6947

C TA G

,.a. A.

B. M 1 2 3 4

target

M 1 2 3 4

I? A 2300 2300

20001353 -

2000-

1078 872603 -

135.31078 872-

I

.C

T T T

CC

-

G

603-

TA

T

..Wemt

A

U3 G

T

G A

G

junction

C. 5'

3'- __U5 PBS

G

integration

G

T

G

C

tat

_~*

I

-3.

U3

R

I

s

MroI/

S

A

-5, T T

435 bp -l- FIG. 5. PCR amplification of autointegration events. (A) Viral DNA purified from infected cell extracts incubated at 37°C for 45 min in either the presence (lanes 1 and 3) or absence (lanes 2 and 4) of 1 mM ATP was amplified by PCR with the PBS and tat primers. After amplification, an aliquot of the reaction mixture was run on a 1% agarose gel either without endonuclease digestion (lanes 1 and 2) or following digestion with MroI (lanes 3 and 4). M, DNA size standards (Hindlll digest of bacteriophage A DNA and HaeIII digest of (X174 DNA). Numbers at left indicate the sizes, in base pairs, of selected fragments. (B) Low-molecular-weight DNA prepared by modified Hirt extraction (6) of uninfected SupTl cells (lanes 2 and 4) or SupTl cells 24 h after infection with HIV-1 (lanes 1 and 3) was amplified by PCR with the PBS and tat primers. Lanes 1 and 2, PCR reaction products without endonuclease digestion; lanes 3 and 4, PCR reaction products following digestion with MroI. Lane M, DNA size standards, as in panel A. (C) Diagrammatic representation of the structure of the products generated by PCR amplification of 2-LTR circles formed by autointegration with the PBS and tat primers. The ends of the amplified DNA fragments are defined by the PBS and tat primers. The site of joining of the viral LTR to target DNA is indicated as the integration junction. The arrow designated sp indicates the approximate location of the sequencing primer used to determine the DNA sequence across the integration junction. The location of the unique MroI site in the U3 region of the LTR is also indicated. (D) DNA sequence analysis of the PCR amplification products generated in the reaction shown in panel A, lane 1. Reaction products were purified from the agarose gel after electrophoresis and sequenced by the method of Sanger et al. (23) with the sequencing primer described in Materials and Methods. The approximate location of this primer is shown in panel C.

DNA sequence analysis provided further evidence that the PCR amplification products resulted from authentic integration events. The heterogeneous products of amplification were purified from agarose gel slices after electrophoresis and sequenced by using a primer that hybridized to the LTR near the U3 terminus. The sequence of the amplified molecules was homogeneous up to the CA dinucleotide present at the end of the U3 region of the LTR; thereafter, the sequence became random (Fig. 5D). This CA dinucleotide is the site of joining of viral DNA to target DNA during integration in vivo (19, 28) and in vitro (8, 20). The randomization of sequence beyond this point results from integration at many target sites within the viral DNA. The identification of the highly conserved CA dinucleotide as the site of LTR joining to target sequence is strong evidence that the recombination events that lead to the formation of 2-LTR circles in vitro are mediated by the viral integrase. PCR analysis was also performed on circular viral DNA formed in vivo, using the primers designed to detect au-

tointegration events. Unintegrated viral DNA was prepared by Hirt extraction of SupTl cells 24 h after infection with HIV-1, at a time when circular viral DNA can be detected in the nucleus (9) (Fig. 2A, lane 5), and subjected to PCR analysis with the primers described above. Again, a smear of DNA fragments was generated by amplification, and a discrete 425-bp band was produced from the amplification products by digestion with MroI (Fig. 5B), indicating that autointegration events occur naturally during viral replication. The formation of authentic 1-LTR circles in vitro was also demonstrated by PCR amplification of reaction products. These experiments were performed with primers in the viral PBS region upstream of the gag gene and in the viral nef gene, each of which directed DNA synthesis toward the respective terminus of the linear DNA molecule (Fig. 1). These primers did not amplify any sequences when viral DNA from untreated extracts was used as a substrate in the PCR reaction (Fig. 6A, lane 1). However, a DNA fragment

6948

FARNET AND HASELTINE

J. VIROL.

A.

B. 1

2

3 4

5 6

1

7 8 9 10 3' 4' 6' 7 8' 9'10 M

-

-

2300 2000

2

3 M

RI

.__ 2300

:

-1353 1078 872

E

--- 603

564

FIG. 6. PCR amplification of in vitro circularization reaction products. (A) Amplification of viral DNA with the PBS and nef primer combination (lanes 1, 3, 5, 7, 9, 3', 7', and 9') or the Rl and R2 primer combination (lanes 2, 4, 6, 8, 10, 4', 6', 8', and 10'). See Fig. 1 for the locations of these primers. All viral DNA substrates were digested with Sail prior to PCR amplification. Following amplification, an aliquot of each reaction mixture was run on a 1% agarose gel either without endonuclease digestion (lanes 1 to 10) or after digestion with MroI (lanes 3' to 10'). Viral DNA substrates used in the PCR were as follows: lanes 1 and 2, deproteinated linear viral DNA incubated at 37°C for 45 min in a cytoplasmic extract of uninfected SupTl cells; lanes 3 and 4, unintegrated viral DNA prepared by Hirt extraction of SupTl cells 24 h after infection with HIV-1; lanes 5 and 6, deproteinated linear viral DNA treated with the Klenow fragment of E. coli DNA polymerase I and T4 DNA ligase; lanes 7 and 8, viral DNA prepared from cytoplasmic extracts of infected cells following incubation of the extracts in the presence of 1 mM ATP at 37°C for 45 min; lanes 9 and 10, deproteinated linear viral DNA incubated at 37°C for 45 min in cytoplasmic extracts of uninfected SupTl cells to which 1 mM ATP had been added. Lanes 3' to 10', MroI digestion of the corresponding reaction mixtures shown in lanes 3 to 10. M, DNA size standards (HindIII digest of phage DNA). Numbers at right indicate the sizes, in base pairs, of selected fragments. (B) PCR amplification of viral DNA prepared from infected cell extracts incubated in the presence of 1 mM ATP for 45 min, using the Rl and R2 primer combination (lanes 1 and 3) or the PBS and nef primer combination (lane 2). Viral DNA substrates were amplified without prior endonuclease digestion (lane 1) or following digestion with Sall and PstI (lanes 2 and 3). Lane M, DNA size standards (in base pairs).

having the size expected for a 1-LTR circle junction was generated when viral DNA prepared from extracts treated with 1 mM ATP was used for amplification (Fig. 6A, lane 7). The amplified product was cut once by the restriction endonuclease MroI, generating fragments with sizes expected of the viral LTR (Fig. 6A, lane 7'). The same product was produced when these primers were used to amplify circular viral DNA prepared from Hirt supernatants of acutely infected cells (Fig. 6A, lane 3), demonstrating that the 1-LTR circles formed in vitro are similar in structure to the 1-LTR circles formed in vivo. Additional rounds of PCR amplification were necessary to detect the presence of a DNA fragment corresponding to the 2-LTR junction fragment when viral DNA from Hirt supernatants of infected cells were amplified with the PBS and nef primers, presumably because of the lower concentration of simple 2-LTR circles compared with 1-LTR circles in these preparations (data not shown). Similarly, 2-LTR junction fragments were not readily detected among the amplification products of circular viral DNA formed in vitro, most likely because of the low concentration of simple 2-LTR circles and competition by the other forms of circular viral DNA (1-LTR circles and 2-LTR circles formed by autointegration) present at much higher concentrations. Therefore, the formation of 2-LTR junctions was analyzed with primers capable of specifically amplifying 2-LTR junction sequences. For these experiments, two primers which anneal to opposite DNA strands in the R region of the viral LTR were used (primers Ri and R2, diagrammed in Fig. 1). These primers

expected to amplify only viral genome-length DNA molecules from linear and 1-LTR circular viral DNA. In addition to the genome-length fragment, simple 2-LTR circles are also predicted to direct the amplification of a small DNA fragment having the length of the viral LTR. Digestion of circular DNA with Sall prior to amplification will prevent the formation of the long DNA molecules but will have no effect on the amplification of the DNA fragment specific to simple 2-LTR circles. Amplification of SalI-digested DNA prepared from Hirt supernatants of acutely infected cells with these primers resulted in the synthesis of the expected LTR-sized DNA fragment (Fig. 6A, lane 4). Note that a weak background smear of amplification products, in addition to the strong fragment corresponding to the 2-LTR junction, was also produced in the PCRs. Since the R region is present twice in 2-LTR circles and the primers anneal to both DNA strands, a population of DNA molecules of heterogeneous lengths is predicted to be produced from 2-LTR circles formed by autointegration when the R-region primers are used, as described earlier for the amplification of autointegration events with the tat and PBS primers (Fig. 1). The weak background of heterogeneous amplification products is consistent with a low but detectable level of autointeare

gration

in vivo.

A DNA fragment having the size expected for the 2-LTR junction was also produced when Sall-digested circular viral DNA formed in vitro in the presence of added ATP was used as a substrate for PCR. However, detection of this fragment was obscured by a high background of amplification prod-

VOL. 65, 1991

CIRCULARIZATION OF HIV-1 DNA IN VITRO

A.

B. 1

2 3 4 5 6

*. §P!! Mo.

1

1

2

2 3 4

3

5 6 7 8 9 10 -

23.1

m

6949

-

23.1

9.4

9.4 6.5 dm

43

FIG. 7. Circularization of viral DNA mediated by preintegration complexes purified by gel filtration chromatography. (A) Cytoplasmic extracts (1-ml volume) were fractionated by Sephacryl S400 gel filtration chromatography as previously described (10). The void volume fractions of the column, containing the peak of viral DNA, were pooled (total volume, 1.5 ml) and incubated as follows: lane 1, column fractions were incubated at 37°C for 45 min; lane 2, linear 4X174 DNA was added to column fractions to a final concentration of 650 ng/ml and incubated at 37°C for 45 min; lane 3, ATP-yS was added to column fractions and incubated at 37°C for 45 min; lane 4, same as lane 3, except that an equal volume of cytoplasmic extract from uninfected SupTl cells was added prior to incubation; lane 5, ATP was added to column fractions to a final concentration of 1 mM and incubated at 37°C for 45 min; lane 6, same as lane 5, except that an equal volume of cytoplasmic extract from uninfected SupTl cells was added prior to incubation. (B) Restriction endonuclease analysis of reaction products. Lanes 1 to 3, products of the reactions shown in panel A, lanes 4 to 6, respectively, digested with Sall and PstI. In both panels, numbers on right show size in kilobases.

ucts of heterogeneous lengths (Fig. 6A, lane 8). Amplification of the specific 2-LTR junction fragment from the products of the in vitro circularization reaction was most likely blocked by the higher concentration of 2-LTR circles formed by autointegration relative to simple 2-LTR circles. As a test of this hypothesis, viral DNA circularized in vitro was first digested with PstI in addition to Sall prior to amplification with the R-region primer. Digestion with these two enzymes is expected to uncouple the LTRs present in those 2-LTR circles produced by autointegration into viral DNA sequences located between the two restriction enzyme recognition sites. Digestion with these enzymes will not interfere with the amplification of the 2-LTR junction present in simple 2-LTR circles. As predicted, pretreatment of the circularization reaction products with PstI and Sall prior to PCR amplification resulted in a reduced background of amplification products and facilitated the detection of the specific 2-LTR junction fragment (Fig. 6B, lane 3). Circularization mediated by partially purified preintegration complexes. The linear viral DNA present in cytoplasmic extracts is a part of a nucleoprotein preintegration complex that can be separated from the bulk of cellular proteins by gel filtration chromatography (2, 8-10). Preintegration complexes were partially purified by Sephacryl S400 column chromatography and tested for the ability to mediate the circularization of viral DNA. Addition of 1 mM ATP to column fractions containing the peak of viral DNA resulted in the formation of only the 2-LTR circular DNA form (Fig. 7A, lane 5). Restriction enzyme analysis indicated that the 2-LTR circular products resulted entirely from autointegration events, as no new 2-LTR junctions were detected (Fig. 7B, lane 2). The addition of cytoplasmic extracts from uninfected SupTl cells to column fractions containing viral DNA restored the ability to form 1-LTR circles (Fig. 7A,

FIG. 8. Circularization of purified linear viral DNA in vitro. Linear viral DNA was purified from cytoplasmic extracts of cells 4 h postinfection as described in Materials and Methods. Purified linear viral DNA was added to cytoplasmic extracts from uninfected SupTl cells and incubated at 37°C for 45 min. The uninfected SupTl cells used for the preparation of extracts were either untreated (lanes 3 to 11) or cultured in the presence of 10 ng of phorbol 12-myristate 13-acetate per ml for 6 h prior to extraction (lane 2). Lane 1 shows the viral DNA forms in extracts of infected cells incubated with 1 mM ATP for 45 min, as size standards. Lanes 2 to 11, reaction products following incubation of purified linear viral DNA in uninfected extracts containing 1 mM ATP (lanes 2 and 3), nothing added (lane 4), 1 mM dATP (lane 5), 1 mM UTP (lane 6), 1 mM CTP (lane 7), 1 mM GTP (lane 8), 1 mM ATP-yS (lane 9), or 1 mM ADP (lane 10). Numbers on right show size in kilobases.

lane 6), indicating that a cellular factor was required for 1-LTR circle formation. In addition, a small amount of new 2-LTR junction fragment was also detected in digests of circular products after the addition of uninfected cell extracts (Fig. 7B, lane 3), indicating that a cellular factor was responsible for generating some of the 2-LTR circular products. Pretreatment of uninfected SupTl extracts with proteinase K prior to addition to the column fractions abolished the ability of the extracts to restore 1-LTR and simple 2-LTR circle formation (data not shown). The formation of all types of circular products depended on the addition of ATP, as viral DNA in column fractions remained linear during incubation at 37°C in the absence of added ATP (Fig. 7A, lane 1), regardless of the addition of uninfected SupTl cell extracts. Addition of ATP-yS to column fractions, in both the presence and absence of uninfected cell extracts, resulted in the exclusive formation of 2-LTR circles by autointegration (Fig. 7A, lanes 3 and 4). This result is consistent with the earlier observation that circular molecules induced by the addition of ATP-yS to unfractionated extracts are entirely the result of autointegration events. Circularization of purified linear viral DNA. Further evidence of a role for host proteins in the formation of the 1-LTR and simple 2-LTR circular forms of viral DNA was obtained by testing the ability of uninfected cell extracts to mediate the circularization of purified viral DNA. Linear viral DNA was prepared from cytoplasmic extracts of uninfected cells by SDS-proteinase K treatment and phenol extraction. The purified DNA was added to cytoplasmic extracts prepared from uninfected SupTl cells in the presence or absence of added ATP. New, slowly migrating DNA species were detected after incubation of linear viral DNA with uninfected cell extracts and added ATP (Fig. 8, lanes 2 and 3). The viral DNA remained linear during incubation in extracts containing no added ATP (Fig. 8, lane 4). The most abundant of the new DNA forms comigrated with the 1-LTR circular DNA molecules formed in infected cell extracts, while a small amount of product migrating as 2-LTR circles could also be detected (Fig. 8, lanes 2 and 3). PCR analysis, using the strategies described above for the amplification of 1-LTR circle junctions and 2-LTR circle junctions, confirmed the formation of authentic 1-LTR and simple 2-LTR circular DNA molecules in these reactions (Fig. 6A, lanes 9 and 10). No amplification products were

6950

FARNET AND HASELTINE

A.

B. 1

2 3 4 5 6

C. 1

2 3 4 5 6 7 4

23.1 -

_0

W,

" %

.-4t

-

9

4

.

.

J. VIROL.

7.

-

"~~~~~~~~~~~:!..

1 2 3

f*23.1 2 9.

.....

9.49 4

FIG. 9. Integration and circularization of viral DNA are competing reactions in vitro. (A) Cytoplasmic extracts from infected cells were incubated at 37°C as follows: lane 1, incubation with 10 pLg of 4fX174 DNA per ml for 45 min; lane 2, incubation with 1 mM ATP for 45 min; lane 3, incubation with 10 jig of 4fX174 DNA per ml and 1 mM ATP for 45 min; lane 4, incubation with 10 jig of XX174 DNA per ml for 45 min and then incubation with 1 mM ATP for 45 min; lane 5, incubation with 1 mM ATP for 45 min and then incubation with 10 jig of X174 DNA per ml for 45 min; lane 6, incubation with nothing added for 45 min and then incubation with 10 ,ug of 4X174 DNA per ml for 45 min. (B) Lanes 1 to 6, cytoplasmic extracts from infected cells were incubated at 37°C for 45 min in the presence of 1 mM ATP and XX174 DNA at a concentration of 0 ,ug/ml (lane 1), 0.3 ,ug/ml (lane 2), 0.8 jig/ml (lane 3), 1.6 ,ug/ml (lane 4), 3.2 jig/ml (lane 5), or 6.4 jig/ml (lane 6). Lane 7, extract incubated with 6.4 ,ug of 4X174 DNA per ml for 45 min in the absence of added ATP. (C) Cytoplasmic extracts from infected cells were incubated at 37°C for 60 min with nothing added (lane 1), with 6.4 ,g of 4)X174 DNA per ml and 1 mM ATP-yS (lane 2), or with 1 mM ATP-yS alone (lane 3). In all panels, numbers on right show size in kilobases.

detected when linear viral DNA incubated in uninfected cell extracts in the absence of added ATP was used as a substrate

for PCR (Fig. 6A, lanes 1 and 2). ATP-yS, which did not induce the formation of 1-LTR or simple 2-LTR circular DNA in cytoplasmic extracts of infected cells, also failed to induce the circularization of purified linear viral DNA when added along with extracts from uninfected cells (Fig. 8A, lane 9). dATP, UTP, and CTP were able to induce the formation of circular DNA when added to uninfected cell extracts containing purified linear viral DNA (Fig. 8, lanes 5 to 7). Under the same conditions, GTP, ddATP, ADP, AMP, and adenosine were unable to induce the formation of detectable amounts of circular viral DNA (Fig. 8, lanes 8 and 10, and data not shown). Competition of integration and circularization reactions. Circularization of the linear viral DNA present in cytoplasmic extracts was competitive with integration into heterologous DNA targets (Fig. 9). Addition of both ATP and target DNA to extracts resulted in the formation of both circular and integrated viral DNA (Fig. 9A, lane 3). Under these conditions, the amounts of circular or integrated reaction products were reduced compared with incubation with ATP alone or target DNA alone (Fig. 9A, lanes 1 and 2). Furthermore, circularization of viral DNA was prevented by preincubation of extracts with target DNA for 30 min before the addition of ATP (Fig. 9A, lane 4). Likewise, integration of viral DNA into target DNA was prevented by preincubation of extracts with ATP prior to the addition of target DNA (Fig. 9A, lane 5). Addition of increasing concentrations of target DNA to extracts containing 1 mM ATP showed that integration of viral DNA into target DNA competed most effectively with the formation of 2-LTR circles (Fig. 9B). Reduced amounts of 2-LTR circles were formed when target DNA was present at 6.4 p,g/ml in extracts containing 1 mM ATP compared with incubation in the presence of ATP and the absence of target DNA (Fig. 9B, lanes 1 and 6). In the presence of a high

concentration of target DNA, the small amount of 2-LTR circles formed was almost entirely simple 2-LTR circles, as the autoradiographic intensity of the 2-LTR circle junction fragment produced by restriction enzyme digestion of reaction products was nearly identical to that of the uncut 2-LTR circles (data not shown). These results indicate that integration into target DNA competed very effectively with the autointegration reaction, such that autointegration 2-LTR circles were not formed in the presence of high concentrations of target DNA. As a further test of this hypothesis, target DNA and ATPyS were added simultaneously to extracts from infected cells (Fig. 9C). Target DNA at a concentration of 6 jig/ml was able to completely inhibit the formation of 2-LTR circles by ATP-yS, circles which were previously shown to result exclusively from autointegration (Fig. 9C, lane 2). Analysis of the formation of 1-LTR circles in the presence of increasing amounts of target DNA indicated that the presence of target DNA did not inhibit 1-LTR circle formation (Fig. 9B). Indeed, in some cases enhanced 1-LTR circle formation was observed when target DNA was added to cell extracts in addition to ATP (for example, see Fig. 9B, lane 5). Addition of very high concentrations (.10 ,g/ml) of target DNA to extracts containing 1 mM ATP reduced the amounts of both 2-LTR and 1-LTR circular DNA formed (Fig. 9A, lane 3). DISCUSSION All the major types of circular viral DNA normally present in the nuclei of infected cells are formed in cytoplasmic extracts of cells newly infected with HIV-1 during incubation of the extracts in the presence of added NTPs. The circular reaction products include 1-LTR circles, 2-LTR circles formed by joining of the ends of the linear DNA, and 2-LTR circles formed by autointegration. This work, together with the previous demonstration that the viral DNA in cytoplasmic extracts can integrate with high efficiency into heterologous target DNA molecules (9), demonstrates that all the major reaction products of viral DNA normally found in infected cells can be made in cell extracts in vitro. Formation of all types of circular viral DNA molecules in vitro requires the addition of NTPs to cell extracts. Adenine NTPs are the most effective inducers of circularization. However, all the NTPs, dNTPs, and ddNTPs tested induced the formation of circular viral DNA. The triphosphate moiety of these compounds is the most important determinant of the ability to induce circularization, as none of the nucleoside diphosphates or monophosphates tested were able to induce the formation of circular DNA. All the NTPs tested were able to induce the formation of 2-LTR circles by autointegration, while only some induced the formation of 1-LTR and simple 2-LTR circles. The formation of circular viral DNA in extracts of cells infected with avian leukosis virus has been described previously (17). It is significant that those experiments were performed with ATP and phosphocreatine and creatine phosphokinase added to the cell extracts, both of which induced the formation of circular viral DNA in the experiments reported here. Autointegration and viral DNA circularization were reported not to occur in extracts from cells infected with murine leukemia virus; however, NTPs were not included in the extracts in those experiments (3). It is possible that the ability of NTPs to induce the formation of circular viral DNA in vitro is a general feature of retroviral preintegration complexes.

VOL. 65, 1991

Cellular factors are required for the formation of 1-LTR and simple 2-LTR circular viral DNAs in vitro. Preintegration complexes purified by gel filtration chromatography supported the formation of 1-LTR and simple 2-LTR circles only after the addition of cytoplasmic extracts from uninfected cells. Furthermore, these two types of circular DNA were formed when deproteinated linear viral DNA was added to cytoplasmic extracts of uninfected cells. These results provide the first direct evidence that host proteins can catalyze the formation of circular retroviral DNA and suggest that the formation of circular viral DNA is independent of viral protein functions. Several lines of evidence suggest that the viral integrase protein is responsible for mediating the autointegration of viral DNA in vitro. PCR amplification and DNA sequence analysis of autointegration reaction products demonstrated that the ends of the linear viral DNA molecule are joined to target sequences at the highly conserved CA dinucleotide characteristic of authentic HIV-1 integration reactions. Preintegration complexes purified by gel filtration chromatography, which contain integrase as the sole viral protein component (10), were able to mediate the autointegration of viral DNA with an efficiency comparable to that of complexes present in unfractionated extracts. Deproteinated viral DNA was not a substrate for autointegration when added to extracts of uninfected cells under conditions which allow the formation of the other circular forms of DNA. Autointegration was effectively blocked by the presence of a high concentration of target DNA during the circularization reaction, suggesting that autointegration and integration into heterologous DNA targets are mediated by the same enzymatic function. With the exception of the 2-LTR circles formed by autointegration, the viral DNA products formed by recombination in vitro were similar in structure to the DNA products formed following transfection of plasmid DNA into mammalian cells (for reviews, see references 1 and 22). Two major types of recombination products are formed when linear plasmid DNA molecules bearing terminal directly repeated sequences are transfected into mammalian cells: full-length circular molecules formed by intramolecular end-to-end joining of the linear plasmid DNA molecule, and circular plasmids containing only a single copy of the formerly repeated sequences (21, 32). Therefore, the host recombination activities responsible for the formation of 1-LTR and simple 2-LTR circular viral DNA may be the same as those responsible for the recombination of transfected DNA molecules. For this reason, the in vitro circularization assays described here may be useful for the biochemical characterization of the functions involved in a wide variety of recombination events in mammalian cells. For example, the formation of 1-LTR circular viral DNA in vitro occurs without the detectable formation of corresponding LTR-sized DNA fragments expected to be produced if recombination results from homologous pairing and strand exchange between the directly repeated LTR sequences (data not shown). This observation is consistent with previously described nonconservative models of homologous recombination (5, 18, 33). According to these models, the LTR sequences at each end of the viral DNA are made single stranded through the action of a processive exonuclease (18) or a DNA-unwinding activity (33), allowing the formation of 1-LTR circles through the intramolecular base pairing of homologous single-stranded termini. For most retroviruses, 1-LTR circles represent the most abundant form of circular viral DNA produced during replication. The circularization of purified linear viral DNA in extracts from uninfected cells provides strong evidence that

CIRCULARIZATION OF HIV-1 DNA IN VITRO

6951

1-LTR circles are the result of host-mediated recombination between the LTRs on linear viral DNA. Alternative models, based on circular DNA molecules produced during the reverse transcription of viral RNA in permeabilized virions, have been proposed to account for the formation of 1-LTR circles (7, 14). In these models, 1-LTR circles result from a circular replication intermediate that results from the failure of reverse transcriptase to catalyze displacement synthesis of the nascent viral DNA through the LTR region. However, the linear viral DNA present in HIV-1 preintegration complexes is full-length, as demonstrated by the ability of restriction endonucleases which cleave only double-stranded DNA to cleave near the termini of linear viral DNA (11). Evidently, full-length linear viral DNA, containing two complete LTRs, is capable of forming 1-LTR circles in vitro, either as a part of the preintegration complex or as purified DNA added to uninfected cell extracts. The role served by NTPs in the circularization of viral DNA is not clear. All the NTPs tested were able to induce the autointegration of viral DNA in preintegration complexes. Hydrolysis of the gamma phosphate is not necessary for the reaction, as ATP-yS induced autointegration at an efficiency comparable to that of ATP. It is possible that NTPs mimic target DNA and induce conformational changes in the preintegration complex necessary for the integration of viral DNA into target DNA. In the absence of heterologous target DNA, NTPs may induce such conformational changes and activate integration into contiguous viral DNA sequences. In contrast to autointegration, 1-LTR and simple 2-LTR circle formation is induced by only a subset of NTPs. A requirement for hydrolysis of the gamma phosphate for the formation of these two types of circles is the most reasonable, but not the only, explanation of the inability of ATP-yS to substitute for ATP in these reactions. Furthermore, the formation of 1-LTR and simple 2-LTR circles is independent of any structural features of the preintegration complex, as deproteinated viral DNA serves as a substrate for the reactions in uninfected cell extracts containing added NTPs. Therefore, the nucleotide requirement for the formation of 1-LTR and simple 2-LTR circles in vitro most likely reflects a substrate requirement of the host recombination functions responsible for the formation of these two types of circles. The relative yields of the circular DNA products formed in vitro depend on the reaction conditions and on the presence or absence of heterologous target DNA. In the absence of added target DNA, approximately equal amounts of 2-LTR and 1-LTR circles are formed in extracts containing 1 mM ATP and 150 mM KCl. Under these conditions, most of the 2-LTR circles result from autointegration events mediated by the viral integrase, while only a small amount of simple 2-LTR circles are detected. Autointegration events are specifically inhibited by adding target DNA to the circularization reactions. In the presence of added target DNA, most of the circular viral DNA formed is of the 1-LTR type. The small amount of 2-LTR circles generated under these conditions results entirely from end-to-end joining of the linear viral DNA. The latter pattern of circular viral DNA, consisting of a large amount of 1-LTR circles and a much smaller amount of simple 2-LTR circles, is characteristic of the relative amounts of circular viral DNA normally found in the nucleus of infected cells (31). Autointegration events disrupt the structural integrity of the viral genome and are likely to be lethal to the virus. The observations presented above suggest that autointegration is inhibited in vivo by the high concentration of target DNA present in the nucleus. No such inhibition of autointegration

6952

FARNET AND HASELTINE

would be expected to occur during the time that the viral preintegration complex is present in the cytoplasm of the cell. Mechanisms must exist in vivo to ensure that the viral

integration machinery is not activated prior to the entry of the preintegration complex into the nucleus. The work presented here indicates that such mechanisms can be overcome under certain conditions in vitro. The in vitro circularization reaction may be useful in identifying compounds capable of stimulating the autointegration of viral DNA before it has entered the nucleus of the cell. Furthermore, the experiments presented here show that the balance of integrated DNA to unintegrated circular DNA depends on the relative rates of competing integration and circularization reactions. Since the replication of HIV-1 and most other retroviruses requires the integration of viral DNA into host DNA, compounds which alter the activity of the enzymes which govern these different processes and which favor circularization over integration may inhibit retroviral replication. ACKNOWLEDGMENTS We thank Dag Helland for many helpful discussions, Emmanuel Zazopoulos for helpful comments on the manuscript, Amy Emmert for the artwork, and Jan Welch for manuscript preparation. This work was supported by grant UOI Al 24845 from the National Institutes of Health National Cooperative Drug Development Program. REFERENCES 1. Bollag, R. J., A. S. Waldman, and R. M. Liskay. 1989. Homologous recombination in mammalian cells. Annu. Rev. Genet. 23:199-225. 2. Bowerman, B., P. 0. Brown, J. M. Bishop, and H. E. Varmus. 1989. A nucleoprotein complex mediates the integration of retroviral DNA. Genes Dev. 3:469-478. 3. Brown, P. O., B. Bowerman, H. E. Varmus, and J. M. Bishop. 1987. Correct integration of retroviral DNA in vitro. Cell 49:347-356. 4. Brown, P. O., B. Bowerman, H. E. Varmus, and J. M. Bishop. 1989. Retroviral integration: structure of the initial covalent product and its precursor, and a role for the IN protein. Proc. Natl. Acad. Sci. USA 86:2525-2529. 5. Chakrabarti, S., and M. M. Seidman. 1986. Intramolecular recombination between transfected repeated sequences in mammalian cells is nonconservative. Mol. Cell. Biol. 6:2520-2526. 6. Chinsky, J., and R. Soeiro. 1982. Studies with aphidicolin on the Fv-1 host restriction of Friend murine leukemia virus. J. Virol. 43:182-190. 7. Dina, D., and E. W. Benz. 1980. Structure of murine sarcoma virus DNA replicative intermediates synthesized in vitro. J. Virol. 33:377-389. 8. Ellison, V., H. Abrams, T. Roe, J. Lifson, and P. Brown. 1990. Human immunodeficiency virus integration in a cell-free system. J. Virol. 64:2711-2715. 9. Farnet, C. M., and W. A. Haseltine. 1990. Integration of human immunodeficiency virus type 1 DNA in vitro. Proc. Natl. Acad. Sci. USA 87:4164-4168. 10. Farnet, C. M., and W. A. Haseltine. 1991. Determination of the viral proteins present in the human immunodeficiency virus type 1 preintegration complex. J. Virol. 65:1910-1915. 11. Farnet, C. M., and W. A. Haseltine. Unpublished data. 12. Fujiwara, T., and K. Mizuuchi. 1988. Retroviral DNA integration: structure of an integration intermediate. Cell 54:497-504. 13. Gilboa, E., S. Goff, A. Shields, F. Yoshimura, S. Mitra, and D. Baltimore. 1979. In vitro synthesis of a 9 kbp terminally redundant DNA carrying the infectivity of Moloney murine leukemia virus. Cell 16:863-874. 14. Junghans, R. P., L. R. Boone, and A. M. Skalka. 1982. Products of reverse transcription in avian retrovirus analyzed by electron microscopy. J. Virol. 43:544-554.

J. VIROL.

15. Katz, R. A., C. A. Omar, J. H. Weis, S. A. Mitsialis, A. J. Faras, and R. V. Guntaka. 1982. Restriction endonuclease and nucleotide sequence analyses of molecularly cloned unintegrated avian tumor virus DNA: structure of large terminal repeats in circle junctions. J. Virol. 42:346-351. 16. Kim, S., R. Byrn, J. Groopman, and D. Baltimore. 1989. Temporal aspects of DNA and RNA synthesis during human immunodeficiency virus infection: evidence for differential gene expression. J. Virol. 63:3708-3713. 17. Lee, Y. M. H., and J. M. Coffin. 1990. Efficient autointegration of avian retrovirus DNA in vitro. J. Virol. 64:5958-5965. 18. Lin, F.-L., K. Sperle, and N. Sternberg. 1984. Model for homologous recombination during transfer of DNA into mouse L cells: role for DNA ends in the recombination process. Mol. Cell. Biol. 4:1020-1034. 19. Muesing, M. A., D. H. Smith, C. D. Cabradilla, C. V. Benton, L. A. Lasky, and D. J. Capon. 1985. Nucleic acid structure and expression of the human AIDS/lymphadenopathy retrovirus. Nature (London) 313:450-458. 20. Myrick, K., C. Farnet, and W. Haseltine. Submitted for publication. 21. Roth, D. B., and J. H. Wilson. 1985. Relative rates of homologous and nonhomologous recombination in transfected DNA. Proc. Natl. Acad. Sci. USA 82:3355-3359. 22. Roth, D. B., and J. H. Wilson. 1988. Illegitimate recombination in mammalian cells, p. 621-653. In R. Kucherlapati and G. R. Smith (ed.), Genetic recombination. American Society for Microbiology, Washington, D.C. 23. Sanger, F., S. Nicklen, and A. R. Coulson. 1977. DNA sequencing with chain-terminating inhibitors. Proc. Natl. Acad. Sci. USA 74:5463-5467. 24. Shank, P. R., S. H. Hughes, H. Kung, J. E. Majors, N. Quintrell, R. V. Guntaka, J. M. Bishop, and H. E. Varmus. 1978. Mapping unintegrated avian sarcoma virus DNA: termini of linear DNA bear 300 nucleotides present once or twice in two species of circular DNA. Cell 15:1383-1395. 25. Shank, P. R., and H. E. Varmus. 1978. Virus-specific DNA in the cytoplasm of avian sarcoma virus-infected cells is a precursor to covalently closed circular viral DNA in the nucleus. J. Virol. 25:104-114. 26. Shoemaker, C., S. Goff, E. Gilboa, M. Paskind, S. W. Mitra, and D. Baltimore. 1980. Structure of a clone of circular Moloney murine leukemia virus DNA molecule containing an inverted segment: implications for retrovirus integration. Proc. Natl. Acad. Sci. USA 77:3732-3736. 27. Shoemaker, C., J. Hoffman, S. P. Goff, and D. Baltimore. 1981. Intramolecular integration within Moloney murine leukemia virus DNA. J. Virol. 40:164-172. 28. Starcich, B., L. Ratner, S. F. Josephs, T. Okamoto, R. C. Gallo, and F. Wong-Staal. 1985. Characterization of long terminal repeat sequences of HTLV-III. Science 227:538-540. 29. Swanstrom, R., W. J. DeLorbe, J. M. Bishop, and H. E. Varmus. 1981. Nucleotide sequence of cloned unintegrated avian sarcoma virus DNA: viral DNA contains direct and inverted repeats similar to those in transposable elements. Proc. Natl. Acad. Sci. USA 78:124-128. 30. Varmus, H. E. 1982. Form and function of retroviral proviruses. Science 216:812-820. 31. Varmus, H. E., and R. Swanstrom. 1982. Replication of retroviruses, p. 369-512. In R. Weiss, N. Teich, H. Varmus, and J. Coffin (ed.), RNA tumor viruses, vol. 1. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 32. Wake, C. T., T. Gudowicz, T. Porter, A. White, and J. H. Wilson. 1984. How damaged is the biologically active subpopulation of transfected DNA? Mol. Cell. Biol. 4:387-398. 33. Wake, C. T., F. Vernaleone, and J. H. Wilson. 1985. Topological requirements for homologous recombination among DNA molecules transfected into mammalian cells. Mol. Cell. Biol. 5:2080-2089. 34. Yoshimura, F. K., and R. A. Weinberg. 1979. Restriction endonuclease cleavage of linear and closed circular murine leukemia DNAs: discovery of a smaller circular form. Cell 16:323-332.