The EMBO Journal vol. 1 1 no.3 pp. 1 131 - 1139, 1992
Domain structure of the human immunodeficiency virus reverse transcriptase
Hermann Lederer1"7, Octavian Schatz2'8, Roland May3, Henry Crespi4, Jean-Luc Darlix5, Stuart F.J.Le Grice6 and Hermann Heumann1'9 1Max Planck Institute of Biochemistry, D-8033 Martinsried, FRG, 2Central Research Units, F.Hoffmann-La Roche Ltd, CH-4002 Basel, Switzerland, 3Institut Laue-Langevin, BP 156, F-38042 Grenoble Cedex 9, France, 4Argonne National Laboratory, IL 60439, USA, 5Ecole Normale Superieure, Laboratoire de Biologie, 46, All6e d'Italie, F-69367 Lyon Cedex 07, France and 6Division of Infectious Diseases, Case Western Reserve University School of Medicine, Cleveland, OH 44106, USA 7Present address: Max Planck Institute of Plasma Physics, Garching, FRG
8Present address: Dept. of Oncology and Virology, Ciba-Geigy Ltd., CH-4002 Basel, Switzerland Communicated by W.Saenger
The spatial arrangement of subunits p51 and p66 of the HIV-1 reverse transcriptase and the position of the RNase H containing domain, p15, have been determined by means of neutron small-angle scattering. The reverse transcriptase (p66/p5i) is a flat molecule, which can be approximated by an ellipsoid with the half axes of 5.2 mm, 4.8 mm and 1.4 mm. The two subunits pSi and p66 having a centre-to-centre distance of 3.3 0.3 nm are attached at their flat sides, slightly shifted sideways. The p15 domain is located at the long axis of the ellipsoidal reverse transcriptase having a distance of 5.0 0.5 am to the centre of the p5ld domain, which is part of the p66 subunit, and a distance of 5.3 + 1.2 mm to the centre of the neighbouring P51s subunit. Key words: HIV-1/neutron scattering/reconstitution/reverse transcriptase/structure
Introduction Reverse transcriptase (RT) catalyses replication of viral genomic RNA. Location of the RNA primer, elongation of the primer by synthesis of a DNA strand complementary to the sequence of the RNA template strand and coordinate digestion of the template are the functional steps mediated by RT, leading to formation of minus strand DNA. Plus strand formation is catalysed by the same enzyme (Varmus and Swanstrom, 1984; Varmus, 1987; Goff, 1990). The functional form of type 1 human immune deficiency virus (HIV-1) is a heterodimer comprising the subunits pSi (molecular weight: 51 kDa) and p66 (molecular weight: 66 kDa; Di Marzo Veronese et al., 1986; Lightfoote et al., 1986; Lowe et al., 1988). The p66 subunit consists of two domains, the p5 Id domain (the subscripts s and d, respectively, denote the subunit and the domain, p51) and the p15 domain. The latter contains the RNase H activity Hansen et al., 1988), whose structure was recently solved Oxford University Press
at atomic resolution (Davies et al., 1991). In this paper we present a model of the solution structure of the isolated HIV- 1 reverse transcriptase at low resolution using neutron small angle scattering. This method provides information about the quaternary structure of biopolymeric complexes as previously shown, e.g. for the Escherichia coli RNA polymerase-DNA complex (Heumann et al., 1988), the tetracycline repressor -DNA complex (Lederer et al., 1989) and the 30S and the 50S ribosomal subunits of E. coli (Capel et al., 1987; Nierhaus et al., 1988). The main advantage of this method is that a single component in a complex can be highlighted by deuteration and the scattering contribution of the remaining part can be matched by contrast variation of the solution buffer with D20 (Crichton et al.,
1977). We applied this technique to the isolated RT and visualized either the p5 1s subunit or the p66 subunit in the heterodimer complex for obtaining the scattering functions of the corresponding subunits in situ. This information, together with the scattering functions of the heterodimer (p66/pS 1) and the homodimer (p66/p66) allowed us to determine the spatial arrangement of the subunits p66 and p51s, and the p15 domain, respectively, within the heterodimeric RT.
Results Scattering parameters of the heterodimeric RT p66/p51 Solutions of free RT with the subunits p66 and p5 1, in a 1:1 stoichiometry were subjected to small angle neutron scattering in H20 (see Materials and methods). These measurements provided the scattering curve of RT (Figure la), from which the corresponding Fourier transform, the distance distribution function, p(r) (Figure lb), was calculated. This function represents the frequency of the distances between any two volume elements of the scattering molecule. An assessment of the quality of the data is often easier by analysing the distance distribution function instead of the original scattering curve. From the distance distribution function the following parameters were derived: (i) the molecular weight (M) of RT, (ii) the maximum dimension (dmax), which is read from the p(r) function at r = rmax, (iii) the radius of gyration (Rg) and the volume (v) (see Materials and methods). These data are compiled in Table I. The parameters d,a, RG and V were used to calculate the shape parameters of RT in terms of a scattering equivalent ellipsoid. According to the procedure described in Materials and methods, an oblate ellipsoid was obtained having the half axes 5.2 nm, 4.8 nm and 1.4 nm. Model calculations have verified that an ellipsoid with these half axes is a good approximation for the shape of the RT (see Figure la). The information about the shape of RT serves as a constraint for the more detailed analysis of the spatial arrangement of the components of RT, the subunits p66 and p5 l, and within the p66 subunit the spatial arrangement of the p5ld domain and the p15 domain. 1131
H.Lederer et al.
b
a
'.4
0.0
1.0
2.0
Q (1/nm)
r(nm)
Fig. 1. (a) The scattering curve of the protonated heterodimeric HIV-1 RT. Filled squares, scattering curve determined experimentally. The error bars lie within the symbols used for data representation. Dashed line, scattering curve calculated for an ellipsoid with the half axes 5.2 nm, 4.8 nm and 1.4 nm. (b) The distance distribution functions, p(r), of the HIV RT components. Filled squares, p(r) of the protonated p5l/p66 heterodimer (RT) in H20. Filled triangles, p(r) of the protonated p66/p66 homodimer in H20. Open triangles, p(r) of the p5lS subunit in situ, obtained from RT (H-D-hybrid) containing a deuterated pSl subunit and a protonated p66 subunit, whose scattering contribution was matched using a 41 % D20 buffer. Open squares, p(r) of the p66 subunit in situ, extrapolated from a set of scattering curves of H-D-hybrid RT at different D20 concentrations (see Figure 3 and Materials and methods).
Scattering parameters of the RT components Determination of the spatial arrangement of the subunits/ domains requires information about the shape of the components, their centre-to-centre distances and their orientations. These parameters can be obtained by neutron scattering, provided a single component or a sub-assembly can be visualized by deuteration. For that purpose we reconstituted an active RT from a deuterated and a protonated (natural abundance hydrogens) subunit.
scattering curve of the p51, subunit in situ, i.e. as a building block of RT. The contrast of the protonated p66 subunit was matched by using a buffer containing 41 % D20 (data not shown). Under these conditions only the deuterated p51 subunit is 'visible' for neutrons. The distance distribution function of the pS5 subunit in situ, which is shown in Figure lb, was obtained by Fourier transformation of the scattering curve. The parameters derived from the distance distribution function are presented in Table I.
Reconstitution of selectively deuterated heterodimer RT. Reconstitution partners were the protonated p66 and deuterated subunits p5l,. The latter had an extension of six histidines at the N-terminus. Due to a dynamic equilibrium between subunits in the dimeric and the monomeric form, a mixture of p66 subunits and p51, subunits contains populations of monomers, homodimers and heterodimers in ratios determined by their equilibrium constants. The yield of heterodimers is high due to the higher affinity constant for the formation of heterodimers (Restle et al., 1991). Figure 2 shows the outline of the reconstitution procedure. Since attempts to reconstitute heterodimer from purified subunits were unsuccessful (S.Le Grice, unpublished) the approach we followed involved co-homogenization of the appropriately induced cultures and metal chelate affinity chromatography (Le Grice et al., 1991). RT reconstituted from protonated p66 and deuterated p51s, the H-D-hybrid contained both subunits in a 1:1 stoichiometry, following chromatography on Ni2+-NTA - and S-Sepharose (Figure 2b). In the present study (data not shown) and related work (Le Grice et al., 1991; Howard et al., 1991), we have demonstrated that reconstitution does not affect the enzymatic properties or tRNALYs 3 binding of the resulting heterodimer (Barat et al., 1989; Darlix, unpublished results).
Shape parameters of the p66 subunit in situ. The scattering curve of the p66 subunit in situ was obtained using the same RT preparation as in the previous experiment. The protonated p66 subunit was visualized after matching the deuterated p5 1 subunit. The contrast of deuterated protein can be eliminated in 1 18% D20. This value of 1 8 1% was obtained by an approach described by Lederer et al. (1986) taking into account the 'zero scattering intensity' (IO) of the H-D-hybrid RT determined at different D20 concentrations (see Figure 3a), as described in Materials and methods. 118% D20 obviously cannot be achieved experimentally. However, an extrapolation procedure described by Stuhrmann permits the determination of the scattering curve of the deuterated p66 subunit from the scattering curves of the H-D-hybrid RT measured at different contrasts (Ibel and Stuhrmann, 1975). We have determined the scattering curves at different D20 concentrations (0, 41, 74 and 92%) and used these curves for obtaining the scattering curve of the complex in 118% D20. Figure 3b illustrates the procedure for obtaining the radius of gyration of the subunit p66 in situ and Figure lb shows the distance distribution function of the p66 subunit obtained according to the Stuhrmann protocol.
Shape parameters of the subunit pSJs in situ. Heterodimeric
RT with a deuterated pSi kDa subunit was subjected to neutron scattering studies with the aim of determining the
1132
7he p15, RNase H-containing domain. The scattering curve of the p15 domain in situ could not be obtained by neutron scattering, since this domain, as part of the p66 subunit, could not be selectively deuterated. The X-ray crystal
Quaternary structure of HIV-1 reverse transcriptase a
D )-[Hisl-p5 1
p66
\
Co-homogenisation 100 000 a g
step I
p66
D-lHis]-p5
p66/D-[His]-p51
b
M
H
N
S c
'3- _---
step
NTA-Seepha-rose metal chc: afffib nity chromatograph:
2
p66/D-[His]-p5 1
66-~~~ al~ ~ ~ ~ ~ ~ ~ ~ ~ *e;-F.: 45-0-
97-2.. 8
mmm
....
._.-p66
_
poo
-[D]
1
D-[His]-p5 1
S-Sepharose ion exchanL chromatographi ste) 3
p66/D-[Hisl-p5
18 -_ s
1
14- am Fig. 2. Reconstitution of p5l/p66 HIV RT containing a deuterated p5IS subunit. (a) Outline of the reconstitution procedure. Cells containing a protonated p66 subunit were mixed with those containing the deuterated pSls subunit (D-[His]pSl) and co-homogenized (step 1). The filled portion at the N-terminus of p51S represents a small polyhistidine extension (His6) to aid purification by metal chelate affinity chromatography (Le Grice and Gruninger-Leitch, 1990). The high speed supematant of a cell homogenate was loaded onto a column of Ni2+-NTA-Sepharose for selection of proteins bearing the polyhistidine extension, i.e., H-D-hybrid RT (D-[His]pS5/H-p66) and D-[His]pSl (step 2). Further chromatography on S-Sepharose separates excess p51 RT from the reconstituted heterodimer (step 3). (b) Analysis of reconstituted, selectively deuterated p5l/p66 HIV-1 RT by SDS gel electrophoresis. Lane M, protein Mr markers; lane H, high speed supernatant; lane N, Ni2+-NTA-Sepharose-purified enzyme; lane D, DEAE-Sephacel-purified enzyme; lane S, S-Sepharose-purified enzyme; lane C, control heterodimer HIV-1 RT. Note that the heterodimer preparation is over-represented by D-[His]pSl until recovery from S-Sepharose.
Table I. Molecular parameters of the components of the HIV- 1 RT Components
p5l p66 p15 p5l/p66 p66/p66 p5l/pSI
Mr (kDa)
Volume
RG
(nm3)
(nm)
51 66 15 117 132 102
63 81 18 145 162
2.48 + 0.04 3.14 0.07 1.4a 3.27 + 0.04 3.53 + 0.04 2.95 0.35b
Dmax (nm)
Half-axes of ellipsoids (nm) a
b
c
8 + 0.5 1 10
4.0 5.0 1.6 5.2 6.0
3.7 3.9
1.0 1.0 1.6 1.4 1.4
10 + 0.5 12 ± 1
1.6 4.8 4.9
The molecular weights obtained by biochemical methods and by neutron scattering agree within 5%. aDetermined from a model of p15 based on the E.coli RNase H (Yang et al., 1990) and moelcular modelling (W.Hendrickson, personal communication).
bCalculated.
structure of p5 has been published recently (Davies et al.,
1991). The
of these data for calculating the distance distribution function of the p15 domain was not possible, since the data were not provided by the authors. Therefore, the radius of gyration (R15) was derived from a model of the p15 domain based on the structure of E.coli RNase H (Yang et al., 1990), which was refined by molecular modelling (W.A.Hendrickson, personal communication). The radius of gyration obtained is R15 = 1.37 nm. A lower limit of R15 = 1.27 nm was estimated assuming a spherical shape of p15 having a partial specific volume of 0.74 cm3/g (Lehninger, 1975). Both values are close, justifying the spherical approximation. The radius r of the corresponding scattering equivalent sphere is 1.6 nm. use
Shape parameters of the homodimer p66/p66. We included in our structural studies also the homodimeric RT (p66/p66). The homodimer differs from the heterodimer by an additional pl5 domain. Both types of RT show almost equivalent DNA polymerizing activity and RNase H activity (Larder et al., 1987; Hizi et al., 1988; Muller et al., 1989; Le Grice and Gruninger-Leitch, 1990). This close enzymatic relatedness encouraged us to compare the structural data obtained from the homodimeric and heterodimeric RT. We overexpressed the homodimeric p66/p66 (Le Grice and Gruninger-Leitch, 1990) and subjected the purified protein to neutron small angle scattering studies. Data evaluation was performed as described above for the heterodimeric RT. The distance distribution function of the p66/p66 homodimer is shown 1133
H.Lederer et at.
2
a
b
12 1 0
._
ce
u-i
8
EC
6
CM
a
4
a2
0
20
60
40
CD20 CH20 C D20
0
--T
Ia-T
-1.5
-1
1 /Ap
-1I
0
-0.5
[10
0.5
1
nmm]
= (ION)/2*signAQ was plotted against the normalized D20 Fig. 3. (a) Contrast variation of the H-D-RT. The normalized scattering amplitude AS concentration, with Io being the normalized 'zero scattering intensity' of the scattering curves determined at the corresponding D20 concentrations, as described in Materials and methods. (b) 'Sturhmann plot' of H-D-hybrid RT for obtaining the distance between the subunits p66 and p5is, and the radius of gyration of the p66 subunit. From a set of scattering curves at different D20 concentrations the radii of gyration (RG) of the H-D-RT were determined by Guinier plots (not shown). The squares of these RC, were plotted against the inverse scattering contrasts 1/AQ and fitted to the parabolic equation (4). From the parameters of this parabola the centre-to-centre distance between the protonated p66 subunit and the deuterated pSls subunit was determined, as shown in Materials and methods. The error bars lie within the size of the symbols used for data representation.
Table 11. Centre-to-centre distances of the components of the HIV-1 RT (a) Calculated from the parallel axes theorem of mechanics Within
Components
P51d
p15
p5ls p66 p5Id/P51s= p15 p66
p66
p51s/p51dd pl5/pl5
p66 as part of the heterodimer p5l/p66 heterodimer p5l/p66 heterodimer p5l/p66 homodimer p66/p66 homodimer p66/p66
zi
Z2
Distance
0.773 0.436 0.872
0.227 0.564 0.128
0.5
0.5
0.773
0.227
5.0 3.25 5.1 3.38 5.1
(nm) ± ± : ± ±
0.5 0.37 1.2 0.42 0.9
Zi are the scattering mass ratios (see Materials and methods).
(b) Calculated by other methods Components
Within
Method
DistanceI
pSi = p66
p66 as part of the heterodimer p5l/p66 heterodimer p5l/p66 heterodimer p5I/p66 heterodimer p51/p66
Stuhrmann plot mass low see Results geometrical construction
3.28 i 0.3 1.14 i 0.11 3.0 + 0.9 5.3 f 1.2
p5ld p51d
p66 pSl
p5Sl
p15
in Figure lb, the radius of gyration and the derived parameters in Table I. Centre-to-centre distances of the RT components
Approximation of the shape of the components by triaxial bodies was the first step in evluating the domain structure of RT. The positioning of these components requires information about their centre-to-centre distances. The neutron scattering approach can give this information provided that the centre of scattering mass determined by neutron scattering coincides with the centre of gravitational mass. This is the case in the approximation applied here, which assumes that subunits have a homogeneous mass distribution. The distance between two components was calculated using three parameters, the radii of gyration of each single component and that of the united components. Assuming that 1134
the scattering parameters obtained for the p5 ls subunit apply also for the pSld domain (which is part of the p66 subunit), the following centre-to-centre distances were calculated by means of the parallel axes theorem of mechanics (see Materials and methods and Table IIa). The distance between the p5ld domain and the p15 domain, d5ld.15, is 5.0 0.5 nm, and the distance between the p51s subunit and the p66 subunit, d5ls-66, is 3.25 il 0.37 nm. This value was verified by an independent distance determination using the 'Stuhrmann plot' (see Materials and methods, Table IIb and Figure 3b) obtaining d5j,_66 value of 3.28 ±E 0.3 nm. The centre-to-centre distance between the p5 Id domain and the p51s subunit in the heterodimer, d5ld-51s, could not be determined directly, since these components could not be deuterated selectively. However, the information about the radius of gyration of the subassembly p5ld/p5ls is contained in the complete p66/pSl,
Quaternary structure of HIV-1 reverse transcriptase
a)
_4
- --_-.-
7.4r Fr
2n
Fig. 4. Reconstruction of the spatial arrangement of the p5ld and the p15, RNase H domain within the p66 subunit. Projection of the ellipsoidal p66 subunit (a) along the two longer axes and (b) along the shortest and the longest axes. 'S' indicates the centre of gravity of the corresponding component. dmax is the maximum dimension of the p66 subunit. The concentric shaded rings indicate the error margins of dmax and of the distance dI5-51d, respectively. The thickest line shows the overall dimension of the p66 subunit represented by a scattering equivalent ellipsoid.
molecule, whose spatial arrangement is known. Using this information an estimation of the distance d5ld-51s is possible. As shown in Materials and methods, we obtained:
2.1 nm < d5ls-51d < 3.9 nm, and for the radius of gyration
R5Sd,51s:
i
S5Id 3
2.6 nm < R5ld,51s < 3.3 nm. I Usin2 this radius of evration. the distance between the centre of the p5ld/p5li sub-assembly and the p15 domain, d5d5 Is -15, was calculated using the parallel axes theorem of mechanics as 5.1 1.2 nm.
Proposed model for p5 1/p66 HIV- 1 RT The centre-to-centre distances of the components in Table II and their shape parameters in Table I were the basis for modelling the domain structure of heterodimeric RT. We started with the spatial arrangement of the components of the p66 subunit, i.e. the p15 domain and the p5Id domain, and then arranged the p51s subunit so that the resulting structure fulfils the distance and shape parameters listed in Tables I and II.
The p15 domain and the pS1d domain within the p66 subunit. Figure 4 shows how the spatial arrangement of the domains of RT was reconstructed. The mass centre of the p15 domain is located on a sphere of radius dI5-5Id = 5.0 0.5 nm around the centre of the p5ld domain. The position of the p5 domain on the sphere is constrained by the maximum dimension of the p66 subunit dmax (p66) = 10 i 1 nm. Figure 4a and b show the geometrical reconstruction in two planes with the longest axis in common. The requirement of contact between the p15- and the p5ld domain restricts the position of p15 to the poles of the ellipsoidal pSld domain. This is obvious in the projection along the longest and shortest axes of the p5ld domain displayed in Figure 4b. It is less obvious in the other projection displayed in Figure 4a. Due to the error margins of the distance parameters, an exact positioning of the p15
Pa\ I
-:unm I v
S51
p5
Fig. 5. Reconstruction of the p5 I/p66 heterodimer. Projection of the heterodimer (a) along the two longest axes and (b) along the two shortest axes but without the p15 domain. Arrows represent the distances between the centres of gravity of the components. The thickest line shows the overall dimension of the p66/pSl RT represented by a scattering equivalent ellipsoid.
domain was not possible. Using in addition the distance ± 0.9 nm) and d51s_15 (5.1 4 0.9 nm), the position of the p15 domain was determined by triangulation. Fixation of the distance triangle within the RT molecule required information about the orientation of the two subunits pSi and p66 within the RT.
d5sld..S (3.0
Spatial arrangement of the p5ls subunit with respect to the p66 subunit. The mass centre of the p51s subunit is located on the distance sphere of radius d66-51s = 3.25 ± 0.37 nm around the centre of the p66 subunit. The position of the 1135
H.Lederer et al.
RNA cutting site
Polymensation site Fig. 6. Quaternary structure of the heterodimeric HIV- 1 RT with the template docked at a putative binding site. The model of RT as obtained from neutron solution scattering was docked to the template, taking into account the following information: (i) the distance between the RNA cutting site and the polymerization site is 15-18 bases (Furfine and Reardon, 1991; Wohrl and Molling, 1990), (ii) the RNA cutting site is located within the p15 domain (this study), (iii) the template binds to the interface between the pSi domain and the p5l subunit (Beccera et al., 1990; Painter et al.,
1990).
p5 1, subunit is further constrained by the overall shape
parameters of the complex p66/pS5s. The shortest half axis of the p66/pS Is complex is only 0.4 nm larger than that of a single p51 component, which is 1.0 nm. This is a strong constraint for the position and the orientation of the p51s subunit on the distance sphere. Both subunits must be arranged side by side on their flat side in order to fit into the overall dimension of the complete RT as shown in Figure Sb. Figure 5a shows a 900 turned projection of the spatial arrangement of the domains of RT. This projection shows that the derived domain structure fits well into the dimension of RT.
Spatial arrangement of the two p66 subunits in the homodimer p66/p66. The distance between the two p66 subunits, d66-66, is 3.38 ±i 0.42 nm (Table Ila). This parameter was calculated using the parallel axes theorem, assuming that the radius of gyration of the single p66 subunit obtained from the heterodimeric RT also holds for the subunits in the homodimer. Concerning the orientation, two arrangements were considered, a parallel alignment of the two p66 subunit in a mirror-symmetrical fashion, which means that the two p15 domains would be positioned in close proximity, and an antiparallel arrangement in a pointsymmetrical fashion, which means that the two p15 domains would be positioned on opposite sides of the homodimer. A distance between the centre of gravity of the p5 Is/pS5 d sub-assembly and the piS/pIS sub-assembly was calculated using the parallel axes theorem. Here the assumption was made that there is no drastic difference between the spatial arrangement of the p5 Is/p5 1d sub-assembly in the heteroand homodimers. A similar value of d51,5Id-15/15 = 5.1 i
1136
0.9 nm was obtained for both the parallel and the antiparallel arrangements (Table Ila). This finding is an independent verification of the corresponding parameter in the heterodimer, the distance d5s515d -15, which was 5.1 ±fi 1.2 nm. Even if data analysis showed that the antiparallel arrangement is more likely, it was not possible to decide which of the two arrangements resides in the homodimer.
Discussion RT is one of the major targets in drug design against AIDS (Jacobo-Molina and Arnold, 1991; Mitsuya et al., 1990; Merluzzi et al., 1990; Pauwels et al., 1990). A better knowledge of the structure and function of this enzyme is essential in the search for suitable drugs. Many models have been proposed in the literature on the basis of speculation and suggestions. Our neutron small angle study is the first approach which provides structural information about RT on the basis of experimental data. A crystallographic approach would have been preferable due to higher resolution, but this approach will have to await the availability of suitable crystals. In contrast to the crystallographic approach, the data interpretation of neutron small angle scattering is less straightforward. Detailed conclusions about the shape would require sophisticated fitting procedures which, in general, do not yield unique results. We avoided this by confining the conclusions about the shapes of the components to ellipsoids, since their half axes can be calculated directly from the experimentally obtained scattering parameters. The strength of neutron small angle scattering is the possibility of gettting information about the distances of the components within a molecule. This
Quaternary structure of HIV-1 reverse transcriptase
information is obtained by solution studies without destroying the molecules. According to the capacity of neutron solution scattering it is the aim of this study to determine the spatial arrangement of the domains of RT rather than their shape. The distance determination between two components is possible if the radii of gyration of these components and their assembly are known. The measurement of the radii of gyration of the components in situ requires selective visualization of the components. This is achieved by isotopic labelling of the single components with deuterium. We succeeded in preparing p5 l/p66 HIV-l RT from a protonated p66 subunit and a deuterated p51s subunit using a novel reconstitution procedure (Figure 2, Le Grice et al., 1991; Howard et al., 1991). Subjection of the hybrid H-D-RT to contrast variation with D20 allowed the determination of the radii of gyration of the p66 and p5l, subunits. The radius of gyration of the p15, RNase H-containing domain was derived from molecular modelling procedures using the E. coli RNase H as a basis (Hendrickson, personal communication, see Results). Using the six radii of gyration of the components p15, p51s and p5ld and the assemblies p51s/p66 (the complete heterodimer), p5ld/P15 (the p66 subunit) and the subassembly, p51d/p51,, the centre-to-centre distances between the components were calculated by the parallel axes theorem. Here the assumption was made that there is no great structural difference between the p51s subunit and the p51d domain in situ, justifying the use of the same radius of gyration for both components. We are aware that minor structural differences between the p5ls subunit and the p51d domain must exist, since both components reside within the RT in a different structural context. However, a drastic conformational change which could affect our conclusions on the RT model is rather unlikely, since the expected structural changes are a consequence of a proteolytic excision of one of the p15 components of the homodimeric p66. The relative error of the distances is 10% . The distance between the p66 and p51s subunits is best determined and verified by an independent approach according to Stuhrmann and Miller (1978). The distance between the p15 domain and the sub-assembly, p51d/p51s, shows the least accuracy. This is due to the small influence of the low molecular weight component, p15, on the radius of gyration of the complex. However, this distance is confirmed by an independent distance determination of these components in the p55/p66 homodimer. Despite this, it could not be decided whether the two p66 subunits are aligned parallel or antiparallel. Therefore, the p5ld domain and the p5ls subunit are displayed in the model in Figure 6 as symmetrical bodies. The centres of gravity of the components are located at the corners of the distance triangle formed by the centre-tocentre distances of the components. The orientation of the components was constrained by the overall shape parameters of the heterodimer. Figure 6 shows that RT is a flat molecule with axes of 10 nm, 9 nm and 3 nm. The two p5i components are positioned side by side, attached at their flat sides. The RNase H domain is positioned at the far end of the ellipsoidal RT, most probably having contacts to both p5l components. Figure 6 shows the model of RT in the complex with its template fitting the nucleic acid template in A-form. This model has to be considered as a tentative model with respect to the position of the nucleic acid. The template is docked -
to RT between the two subunits p66 and p5 1, as proposed previously (Jacobo-Molina and Arnold, 1991; Arnold and Arnold, 1991). This position was proposed, taking into account the following information: (i) the functionally active form of HIV RT is the dimeric form (Restle et al., 1990) although the p66 subunit alone contains all functional activities (Cheng et al., 1991; Wu et al., 1991), (ii) there is only one template binding site on the heterodimeric RT (Painter et al., 1990) and (iii) the two subunits p5Is and p66 interact via a hydrophobic region (Becerra et al., 1990; Painter et al., 1990), which might serve as the template binding site. The orientation of RT shown in Figure 6 is based on our knowledge that the RNA template in a DNA -RNA hybrid is cut upstream of the polymerization site (Schatz et al., 1990; Furfine and Reardon, 1991). The number of bases between the RNA cutting site and the polymerization site varies between 7 and 18 bases (Oyama et al., 1989; Schatz et al., 1990; Furfine and Reardon, 1991; Wohrl and Molling, 1990). We used the greatest value of 16-18 bases in order to map the polymerization site in our model.
Conclusion Our neutron scattering studies on selectively deuterated p5I/p66 HIV-1 RT provide information about the spatial arrangement of the components of RT, i.e. the subunit p51s and the domains p51d and p15, in solution. The model derived is considered as a working model, which can serve as a basis for designing new experiments concerning RT structure and function e.g. (i) the role of the p51 subunit in comparison to the p66 subunit, which carries all functional activities of RT, (ii) the conformational change induced in the template following RT binding, (iii) the portion of the template which interacts with RT. Answers to these questions might also be significant for the design of a nucleic acid fragment suitable for co-crystallization of RT with its template.
Materials and methods Preparation of RT Heterodimeric HIV-1 RT was prepared by metal chelate affinity chromatography from E coli strain M 15:: pDMI. 1:: pRT6H-PROT, which directs expression of p66 RT and HIV- 1 protease (Le Grice and GruningerLeitch, 1990). Enzyme thus prepared bears a polyhistidine extension on the C-terminus of the 66 kDa subunit. Reconstituted, selectively deuterated HIV-1 RT was prepared according to published procedures (Le Grice et al., 1991), with the exception that (i) bacteria expressing His-pSI (Schatz et al., 1990) were grown and induced in deuterated algal hydrolysate and (ii) DEAE-Sephacel and S-Sepharose chromatography followed
Ni2+-NTA-Sepharose purification. The protocol for preparation of this enzyme, D-[HisIpSl/Hp66, is outlined schematically in Figure 2a. Homodimeric p66/p66 R was prepared from the strain M 15:: pDM1. 1:: p6HRT as previously described (Le Grice and Gruninger-Leitch, 1990). All enzyme preparations were judged by SDS electrophoresis to be at least 95% homogeneous. Growth and induction of the recombinant E. coli strains was as previously described (Le Grice et al., 1991). In the experiments reported here and related studies (Le Grice et al., 1991; Howard et al., 1991), p5l/p66 prepared by reconstitution retained full polymerase and RNase H activities. Furthermore, tRNALYS.3 binding to the reconstituted, selectively deuterated enzyme was uanffected (J.-L.Darlix, unpublished). The concentration of each protein was determined using an extinction coefficient e1% of 18 for both the homodimer and heterodimer. Data collection The small-angle neutron scattering measurements were performed on the DII small-angle diffractometer of the Institut Laue-Langevin (Grenoble,
1137
H.Lederer et al. France). The samples were dialysed against buffers containing 0.05 M or 0.5 M NaCl, pH 7.9, 10 mM 13-mercaptoehanol in H20, D20 or mixtures of both. During beam exposure the samples were equilibrated to 8°C. The scattering data were processed and corrected as described by Heumann et al. (1988). Concentration effects were eliminated by linearly extrapolating the scattering curves, which were measured at two different concentrations, usually 6 mg/mi and 3 mg/ml, to concentration zero. From the scattering curves the distance distribution functions p(r) were calculated [with a Fortran program of Glatter (1977)], from which the maximum dimensions of the particles were obtained. Forward scattered intensities and radii of gyration were determined from the area and second moment of the p(r) functions, as well as from Guinier plots of ln[I(Q)] versus Q2, where Q is the momentum transfer,
Q = (4) X sin e x where 20 is the full scattering angle. The half axes (a, b, and c) of a scattering equivalent ellipsoid were calculated from the radius of gyration RG and the maximum dimension dmax = 2c using the following equations: R 2 = a2 + h2 + ((1) and
V= 43 r X a x b x c
(2)
The volume (V) of a protein component was calculated from its molecular weight assuming a partial specific volume of v = 0.74 cm3/g (Lehninger, 1975). Distance determination Parallel axes theorem. The centre-to-centre distance d, 2 of two components with the radii of gyration RI and R2 was determined from the radius of gyration R112 of the complexed particle according to the parallel axes theorem of mechanics applied to scattering: Rj,2 =
zi
x
RI
+Z2 z2 -
x Z2 x
dl2
with RC = 10.93 nm, the radius of gyration at infinite contrast, and the coefficients a = -2.58 x 10-4 and 3 = 1.31 x 10-7 nm-2. The quality of the fit is indicated by the correlation coefficient of 0.999. Using the coefficient ( the distance between the components H-p66 and D-pSI is obtained according to the equation
dsl,
- 66 = (
AQi
)
+
( Q2
2
(5)
)
AQ and AQ2 are the contrasts at which the subunits H-p66 and D-p51 matched, respectively. The centre-to-centre distance obtained is
d5l,566
are
= 3.28 + 0.3 nm.
Extrapolation of the plot to the contrast matching of deuterated p51 in a buffer containing the hypothetical D20 concentraton of 118% D20, yielded the radius of gyration, R66 of 3.14 + 0.07 nm. The mass law. Within the p66 subunit, the distance d66-51d between the centres of the component p66 and the p5ld is correlated with the known distance, d66-15, between p66 and p15 according to their molecular weights
M51
and
M15
as
follows:
d66-51d x
M51
=
d66-15
x
M15
(6)
Estimation of the distance between the centres of the pSJd domain and the pSI, subunit, and of the radius of gyration Rs5d/s,5. The centre of the p5 1d domain lies on the sphere of radius d51d-66 = 1.1 0.1 nm centred around the p66 subunit. This distance is calculated according to the mass law (6). An upper and lower limit of this distance is therefore, d515_51d = d5is _66 d5ld-66 = 3.3 i 1.2 nm. The upper limit, (d5l 51d))max .. is 4.5 nm. This can be further restricted to a value of 3.9 nm using the parallel axes theorem (3) by taking into account that an upper limit of the radius of gyration of the p5ld/p5l, sub-assembly is the radius of gyration of the complete heterodimer, R66/51s = 3.27 + 0.04 nm. It then follows that: 2.1 nm
Domain structure of the human immunodeficiency virus reverse transcriptase
Hermann Lederer1"7, Octavian Schatz2'8, Roland May3, Henry Crespi4, Jean-Luc Darlix5, Stuart F.J.Le Grice6 and Hermann Heumann1'9 1Max Planck Institute of Biochemistry, D-8033 Martinsried, FRG, 2Central Research Units, F.Hoffmann-La Roche Ltd, CH-4002 Basel, Switzerland, 3Institut Laue-Langevin, BP 156, F-38042 Grenoble Cedex 9, France, 4Argonne National Laboratory, IL 60439, USA, 5Ecole Normale Superieure, Laboratoire de Biologie, 46, All6e d'Italie, F-69367 Lyon Cedex 07, France and 6Division of Infectious Diseases, Case Western Reserve University School of Medicine, Cleveland, OH 44106, USA 7Present address: Max Planck Institute of Plasma Physics, Garching, FRG
8Present address: Dept. of Oncology and Virology, Ciba-Geigy Ltd., CH-4002 Basel, Switzerland Communicated by W.Saenger
The spatial arrangement of subunits p51 and p66 of the HIV-1 reverse transcriptase and the position of the RNase H containing domain, p15, have been determined by means of neutron small-angle scattering. The reverse transcriptase (p66/p5i) is a flat molecule, which can be approximated by an ellipsoid with the half axes of 5.2 mm, 4.8 mm and 1.4 mm. The two subunits pSi and p66 having a centre-to-centre distance of 3.3 0.3 nm are attached at their flat sides, slightly shifted sideways. The p15 domain is located at the long axis of the ellipsoidal reverse transcriptase having a distance of 5.0 0.5 am to the centre of the p5ld domain, which is part of the p66 subunit, and a distance of 5.3 + 1.2 mm to the centre of the neighbouring P51s subunit. Key words: HIV-1/neutron scattering/reconstitution/reverse transcriptase/structure
Introduction Reverse transcriptase (RT) catalyses replication of viral genomic RNA. Location of the RNA primer, elongation of the primer by synthesis of a DNA strand complementary to the sequence of the RNA template strand and coordinate digestion of the template are the functional steps mediated by RT, leading to formation of minus strand DNA. Plus strand formation is catalysed by the same enzyme (Varmus and Swanstrom, 1984; Varmus, 1987; Goff, 1990). The functional form of type 1 human immune deficiency virus (HIV-1) is a heterodimer comprising the subunits pSi (molecular weight: 51 kDa) and p66 (molecular weight: 66 kDa; Di Marzo Veronese et al., 1986; Lightfoote et al., 1986; Lowe et al., 1988). The p66 subunit consists of two domains, the p5 Id domain (the subscripts s and d, respectively, denote the subunit and the domain, p51) and the p15 domain. The latter contains the RNase H activity Hansen et al., 1988), whose structure was recently solved Oxford University Press
at atomic resolution (Davies et al., 1991). In this paper we present a model of the solution structure of the isolated HIV- 1 reverse transcriptase at low resolution using neutron small angle scattering. This method provides information about the quaternary structure of biopolymeric complexes as previously shown, e.g. for the Escherichia coli RNA polymerase-DNA complex (Heumann et al., 1988), the tetracycline repressor -DNA complex (Lederer et al., 1989) and the 30S and the 50S ribosomal subunits of E. coli (Capel et al., 1987; Nierhaus et al., 1988). The main advantage of this method is that a single component in a complex can be highlighted by deuteration and the scattering contribution of the remaining part can be matched by contrast variation of the solution buffer with D20 (Crichton et al.,
1977). We applied this technique to the isolated RT and visualized either the p5 1s subunit or the p66 subunit in the heterodimer complex for obtaining the scattering functions of the corresponding subunits in situ. This information, together with the scattering functions of the heterodimer (p66/pS 1) and the homodimer (p66/p66) allowed us to determine the spatial arrangement of the subunits p66 and p51s, and the p15 domain, respectively, within the heterodimeric RT.
Results Scattering parameters of the heterodimeric RT p66/p51 Solutions of free RT with the subunits p66 and p5 1, in a 1:1 stoichiometry were subjected to small angle neutron scattering in H20 (see Materials and methods). These measurements provided the scattering curve of RT (Figure la), from which the corresponding Fourier transform, the distance distribution function, p(r) (Figure lb), was calculated. This function represents the frequency of the distances between any two volume elements of the scattering molecule. An assessment of the quality of the data is often easier by analysing the distance distribution function instead of the original scattering curve. From the distance distribution function the following parameters were derived: (i) the molecular weight (M) of RT, (ii) the maximum dimension (dmax), which is read from the p(r) function at r = rmax, (iii) the radius of gyration (Rg) and the volume (v) (see Materials and methods). These data are compiled in Table I. The parameters d,a, RG and V were used to calculate the shape parameters of RT in terms of a scattering equivalent ellipsoid. According to the procedure described in Materials and methods, an oblate ellipsoid was obtained having the half axes 5.2 nm, 4.8 nm and 1.4 nm. Model calculations have verified that an ellipsoid with these half axes is a good approximation for the shape of the RT (see Figure la). The information about the shape of RT serves as a constraint for the more detailed analysis of the spatial arrangement of the components of RT, the subunits p66 and p5 l, and within the p66 subunit the spatial arrangement of the p5ld domain and the p15 domain. 1131
H.Lederer et al.
b
a
'.4
0.0
1.0
2.0
Q (1/nm)
r(nm)
Fig. 1. (a) The scattering curve of the protonated heterodimeric HIV-1 RT. Filled squares, scattering curve determined experimentally. The error bars lie within the symbols used for data representation. Dashed line, scattering curve calculated for an ellipsoid with the half axes 5.2 nm, 4.8 nm and 1.4 nm. (b) The distance distribution functions, p(r), of the HIV RT components. Filled squares, p(r) of the protonated p5l/p66 heterodimer (RT) in H20. Filled triangles, p(r) of the protonated p66/p66 homodimer in H20. Open triangles, p(r) of the p5lS subunit in situ, obtained from RT (H-D-hybrid) containing a deuterated pSl subunit and a protonated p66 subunit, whose scattering contribution was matched using a 41 % D20 buffer. Open squares, p(r) of the p66 subunit in situ, extrapolated from a set of scattering curves of H-D-hybrid RT at different D20 concentrations (see Figure 3 and Materials and methods).
Scattering parameters of the RT components Determination of the spatial arrangement of the subunits/ domains requires information about the shape of the components, their centre-to-centre distances and their orientations. These parameters can be obtained by neutron scattering, provided a single component or a sub-assembly can be visualized by deuteration. For that purpose we reconstituted an active RT from a deuterated and a protonated (natural abundance hydrogens) subunit.
scattering curve of the p51, subunit in situ, i.e. as a building block of RT. The contrast of the protonated p66 subunit was matched by using a buffer containing 41 % D20 (data not shown). Under these conditions only the deuterated p51 subunit is 'visible' for neutrons. The distance distribution function of the pS5 subunit in situ, which is shown in Figure lb, was obtained by Fourier transformation of the scattering curve. The parameters derived from the distance distribution function are presented in Table I.
Reconstitution of selectively deuterated heterodimer RT. Reconstitution partners were the protonated p66 and deuterated subunits p5l,. The latter had an extension of six histidines at the N-terminus. Due to a dynamic equilibrium between subunits in the dimeric and the monomeric form, a mixture of p66 subunits and p51, subunits contains populations of monomers, homodimers and heterodimers in ratios determined by their equilibrium constants. The yield of heterodimers is high due to the higher affinity constant for the formation of heterodimers (Restle et al., 1991). Figure 2 shows the outline of the reconstitution procedure. Since attempts to reconstitute heterodimer from purified subunits were unsuccessful (S.Le Grice, unpublished) the approach we followed involved co-homogenization of the appropriately induced cultures and metal chelate affinity chromatography (Le Grice et al., 1991). RT reconstituted from protonated p66 and deuterated p51s, the H-D-hybrid contained both subunits in a 1:1 stoichiometry, following chromatography on Ni2+-NTA - and S-Sepharose (Figure 2b). In the present study (data not shown) and related work (Le Grice et al., 1991; Howard et al., 1991), we have demonstrated that reconstitution does not affect the enzymatic properties or tRNALYs 3 binding of the resulting heterodimer (Barat et al., 1989; Darlix, unpublished results).
Shape parameters of the p66 subunit in situ. The scattering curve of the p66 subunit in situ was obtained using the same RT preparation as in the previous experiment. The protonated p66 subunit was visualized after matching the deuterated p5 1 subunit. The contrast of deuterated protein can be eliminated in 1 18% D20. This value of 1 8 1% was obtained by an approach described by Lederer et al. (1986) taking into account the 'zero scattering intensity' (IO) of the H-D-hybrid RT determined at different D20 concentrations (see Figure 3a), as described in Materials and methods. 118% D20 obviously cannot be achieved experimentally. However, an extrapolation procedure described by Stuhrmann permits the determination of the scattering curve of the deuterated p66 subunit from the scattering curves of the H-D-hybrid RT measured at different contrasts (Ibel and Stuhrmann, 1975). We have determined the scattering curves at different D20 concentrations (0, 41, 74 and 92%) and used these curves for obtaining the scattering curve of the complex in 118% D20. Figure 3b illustrates the procedure for obtaining the radius of gyration of the subunit p66 in situ and Figure lb shows the distance distribution function of the p66 subunit obtained according to the Stuhrmann protocol.
Shape parameters of the subunit pSJs in situ. Heterodimeric
RT with a deuterated pSi kDa subunit was subjected to neutron scattering studies with the aim of determining the
1132
7he p15, RNase H-containing domain. The scattering curve of the p15 domain in situ could not be obtained by neutron scattering, since this domain, as part of the p66 subunit, could not be selectively deuterated. The X-ray crystal
Quaternary structure of HIV-1 reverse transcriptase a
D )-[Hisl-p5 1
p66
\
Co-homogenisation 100 000 a g
step I
p66
D-lHis]-p5
p66/D-[His]-p51
b
M
H
N
S c
'3- _---
step
NTA-Seepha-rose metal chc: afffib nity chromatograph:
2
p66/D-[His]-p5 1
66-~~~ al~ ~ ~ ~ ~ ~ ~ ~ ~ *e;-F.: 45-0-
97-2.. 8
mmm
....
._.-p66
_
poo
-[D]
1
D-[His]-p5 1
S-Sepharose ion exchanL chromatographi ste) 3
p66/D-[Hisl-p5
18 -_ s
1
14- am Fig. 2. Reconstitution of p5l/p66 HIV RT containing a deuterated p5IS subunit. (a) Outline of the reconstitution procedure. Cells containing a protonated p66 subunit were mixed with those containing the deuterated pSls subunit (D-[His]pSl) and co-homogenized (step 1). The filled portion at the N-terminus of p51S represents a small polyhistidine extension (His6) to aid purification by metal chelate affinity chromatography (Le Grice and Gruninger-Leitch, 1990). The high speed supematant of a cell homogenate was loaded onto a column of Ni2+-NTA-Sepharose for selection of proteins bearing the polyhistidine extension, i.e., H-D-hybrid RT (D-[His]pS5/H-p66) and D-[His]pSl (step 2). Further chromatography on S-Sepharose separates excess p51 RT from the reconstituted heterodimer (step 3). (b) Analysis of reconstituted, selectively deuterated p5l/p66 HIV-1 RT by SDS gel electrophoresis. Lane M, protein Mr markers; lane H, high speed supernatant; lane N, Ni2+-NTA-Sepharose-purified enzyme; lane D, DEAE-Sephacel-purified enzyme; lane S, S-Sepharose-purified enzyme; lane C, control heterodimer HIV-1 RT. Note that the heterodimer preparation is over-represented by D-[His]pSl until recovery from S-Sepharose.
Table I. Molecular parameters of the components of the HIV- 1 RT Components
p5l p66 p15 p5l/p66 p66/p66 p5l/pSI
Mr (kDa)
Volume
RG
(nm3)
(nm)
51 66 15 117 132 102
63 81 18 145 162
2.48 + 0.04 3.14 0.07 1.4a 3.27 + 0.04 3.53 + 0.04 2.95 0.35b
Dmax (nm)
Half-axes of ellipsoids (nm) a
b
c
8 + 0.5 1 10
4.0 5.0 1.6 5.2 6.0
3.7 3.9
1.0 1.0 1.6 1.4 1.4
10 + 0.5 12 ± 1
1.6 4.8 4.9
The molecular weights obtained by biochemical methods and by neutron scattering agree within 5%. aDetermined from a model of p15 based on the E.coli RNase H (Yang et al., 1990) and moelcular modelling (W.Hendrickson, personal communication).
bCalculated.
structure of p5 has been published recently (Davies et al.,
1991). The
of these data for calculating the distance distribution function of the p15 domain was not possible, since the data were not provided by the authors. Therefore, the radius of gyration (R15) was derived from a model of the p15 domain based on the structure of E.coli RNase H (Yang et al., 1990), which was refined by molecular modelling (W.A.Hendrickson, personal communication). The radius of gyration obtained is R15 = 1.37 nm. A lower limit of R15 = 1.27 nm was estimated assuming a spherical shape of p15 having a partial specific volume of 0.74 cm3/g (Lehninger, 1975). Both values are close, justifying the spherical approximation. The radius r of the corresponding scattering equivalent sphere is 1.6 nm. use
Shape parameters of the homodimer p66/p66. We included in our structural studies also the homodimeric RT (p66/p66). The homodimer differs from the heterodimer by an additional pl5 domain. Both types of RT show almost equivalent DNA polymerizing activity and RNase H activity (Larder et al., 1987; Hizi et al., 1988; Muller et al., 1989; Le Grice and Gruninger-Leitch, 1990). This close enzymatic relatedness encouraged us to compare the structural data obtained from the homodimeric and heterodimeric RT. We overexpressed the homodimeric p66/p66 (Le Grice and Gruninger-Leitch, 1990) and subjected the purified protein to neutron small angle scattering studies. Data evaluation was performed as described above for the heterodimeric RT. The distance distribution function of the p66/p66 homodimer is shown 1133
H.Lederer et at.
2
a
b
12 1 0
._
ce
u-i
8
EC
6
CM
a
4
a2
0
20
60
40
CD20 CH20 C D20
0
--T
Ia-T
-1.5
-1
1 /Ap
-1I
0
-0.5
[10
0.5
1
nmm]
= (ION)/2*signAQ was plotted against the normalized D20 Fig. 3. (a) Contrast variation of the H-D-RT. The normalized scattering amplitude AS concentration, with Io being the normalized 'zero scattering intensity' of the scattering curves determined at the corresponding D20 concentrations, as described in Materials and methods. (b) 'Sturhmann plot' of H-D-hybrid RT for obtaining the distance between the subunits p66 and p5is, and the radius of gyration of the p66 subunit. From a set of scattering curves at different D20 concentrations the radii of gyration (RG) of the H-D-RT were determined by Guinier plots (not shown). The squares of these RC, were plotted against the inverse scattering contrasts 1/AQ and fitted to the parabolic equation (4). From the parameters of this parabola the centre-to-centre distance between the protonated p66 subunit and the deuterated pSls subunit was determined, as shown in Materials and methods. The error bars lie within the size of the symbols used for data representation.
Table 11. Centre-to-centre distances of the components of the HIV-1 RT (a) Calculated from the parallel axes theorem of mechanics Within
Components
P51d
p15
p5ls p66 p5Id/P51s= p15 p66
p66
p51s/p51dd pl5/pl5
p66 as part of the heterodimer p5l/p66 heterodimer p5l/p66 heterodimer p5l/p66 homodimer p66/p66 homodimer p66/p66
zi
Z2
Distance
0.773 0.436 0.872
0.227 0.564 0.128
0.5
0.5
0.773
0.227
5.0 3.25 5.1 3.38 5.1
(nm) ± ± : ± ±
0.5 0.37 1.2 0.42 0.9
Zi are the scattering mass ratios (see Materials and methods).
(b) Calculated by other methods Components
Within
Method
DistanceI
pSi = p66
p66 as part of the heterodimer p5l/p66 heterodimer p5l/p66 heterodimer p5I/p66 heterodimer p51/p66
Stuhrmann plot mass low see Results geometrical construction
3.28 i 0.3 1.14 i 0.11 3.0 + 0.9 5.3 f 1.2
p5ld p51d
p66 pSl
p5Sl
p15
in Figure lb, the radius of gyration and the derived parameters in Table I. Centre-to-centre distances of the RT components
Approximation of the shape of the components by triaxial bodies was the first step in evluating the domain structure of RT. The positioning of these components requires information about their centre-to-centre distances. The neutron scattering approach can give this information provided that the centre of scattering mass determined by neutron scattering coincides with the centre of gravitational mass. This is the case in the approximation applied here, which assumes that subunits have a homogeneous mass distribution. The distance between two components was calculated using three parameters, the radii of gyration of each single component and that of the united components. Assuming that 1134
the scattering parameters obtained for the p5 ls subunit apply also for the pSld domain (which is part of the p66 subunit), the following centre-to-centre distances were calculated by means of the parallel axes theorem of mechanics (see Materials and methods and Table IIa). The distance between the p5ld domain and the p15 domain, d5ld.15, is 5.0 0.5 nm, and the distance between the p51s subunit and the p66 subunit, d5ls-66, is 3.25 il 0.37 nm. This value was verified by an independent distance determination using the 'Stuhrmann plot' (see Materials and methods, Table IIb and Figure 3b) obtaining d5j,_66 value of 3.28 ±E 0.3 nm. The centre-to-centre distance between the p5 Id domain and the p51s subunit in the heterodimer, d5ld-51s, could not be determined directly, since these components could not be deuterated selectively. However, the information about the radius of gyration of the subassembly p5ld/p5ls is contained in the complete p66/pSl,
Quaternary structure of HIV-1 reverse transcriptase
a)
_4
- --_-.-
7.4r Fr
2n
Fig. 4. Reconstruction of the spatial arrangement of the p5ld and the p15, RNase H domain within the p66 subunit. Projection of the ellipsoidal p66 subunit (a) along the two longer axes and (b) along the shortest and the longest axes. 'S' indicates the centre of gravity of the corresponding component. dmax is the maximum dimension of the p66 subunit. The concentric shaded rings indicate the error margins of dmax and of the distance dI5-51d, respectively. The thickest line shows the overall dimension of the p66 subunit represented by a scattering equivalent ellipsoid.
molecule, whose spatial arrangement is known. Using this information an estimation of the distance d5ld-51s is possible. As shown in Materials and methods, we obtained:
2.1 nm < d5ls-51d < 3.9 nm, and for the radius of gyration
R5Sd,51s:
i
S5Id 3
2.6 nm < R5ld,51s < 3.3 nm. I Usin2 this radius of evration. the distance between the centre of the p5ld/p5li sub-assembly and the p15 domain, d5d5 Is -15, was calculated using the parallel axes theorem of mechanics as 5.1 1.2 nm.
Proposed model for p5 1/p66 HIV- 1 RT The centre-to-centre distances of the components in Table II and their shape parameters in Table I were the basis for modelling the domain structure of heterodimeric RT. We started with the spatial arrangement of the components of the p66 subunit, i.e. the p15 domain and the p5Id domain, and then arranged the p51s subunit so that the resulting structure fulfils the distance and shape parameters listed in Tables I and II.
The p15 domain and the pS1d domain within the p66 subunit. Figure 4 shows how the spatial arrangement of the domains of RT was reconstructed. The mass centre of the p15 domain is located on a sphere of radius dI5-5Id = 5.0 0.5 nm around the centre of the p5ld domain. The position of the p5 domain on the sphere is constrained by the maximum dimension of the p66 subunit dmax (p66) = 10 i 1 nm. Figure 4a and b show the geometrical reconstruction in two planes with the longest axis in common. The requirement of contact between the p15- and the p5ld domain restricts the position of p15 to the poles of the ellipsoidal pSld domain. This is obvious in the projection along the longest and shortest axes of the p5ld domain displayed in Figure 4b. It is less obvious in the other projection displayed in Figure 4a. Due to the error margins of the distance parameters, an exact positioning of the p15
Pa\ I
-:unm I v
S51
p5
Fig. 5. Reconstruction of the p5 I/p66 heterodimer. Projection of the heterodimer (a) along the two longest axes and (b) along the two shortest axes but without the p15 domain. Arrows represent the distances between the centres of gravity of the components. The thickest line shows the overall dimension of the p66/pSl RT represented by a scattering equivalent ellipsoid.
domain was not possible. Using in addition the distance ± 0.9 nm) and d51s_15 (5.1 4 0.9 nm), the position of the p15 domain was determined by triangulation. Fixation of the distance triangle within the RT molecule required information about the orientation of the two subunits pSi and p66 within the RT.
d5sld..S (3.0
Spatial arrangement of the p5ls subunit with respect to the p66 subunit. The mass centre of the p51s subunit is located on the distance sphere of radius d66-51s = 3.25 ± 0.37 nm around the centre of the p66 subunit. The position of the 1135
H.Lederer et al.
RNA cutting site
Polymensation site Fig. 6. Quaternary structure of the heterodimeric HIV- 1 RT with the template docked at a putative binding site. The model of RT as obtained from neutron solution scattering was docked to the template, taking into account the following information: (i) the distance between the RNA cutting site and the polymerization site is 15-18 bases (Furfine and Reardon, 1991; Wohrl and Molling, 1990), (ii) the RNA cutting site is located within the p15 domain (this study), (iii) the template binds to the interface between the pSi domain and the p5l subunit (Beccera et al., 1990; Painter et al.,
1990).
p5 1, subunit is further constrained by the overall shape
parameters of the complex p66/pS5s. The shortest half axis of the p66/pS Is complex is only 0.4 nm larger than that of a single p51 component, which is 1.0 nm. This is a strong constraint for the position and the orientation of the p51s subunit on the distance sphere. Both subunits must be arranged side by side on their flat side in order to fit into the overall dimension of the complete RT as shown in Figure Sb. Figure 5a shows a 900 turned projection of the spatial arrangement of the domains of RT. This projection shows that the derived domain structure fits well into the dimension of RT.
Spatial arrangement of the two p66 subunits in the homodimer p66/p66. The distance between the two p66 subunits, d66-66, is 3.38 ±i 0.42 nm (Table Ila). This parameter was calculated using the parallel axes theorem, assuming that the radius of gyration of the single p66 subunit obtained from the heterodimeric RT also holds for the subunits in the homodimer. Concerning the orientation, two arrangements were considered, a parallel alignment of the two p66 subunit in a mirror-symmetrical fashion, which means that the two p15 domains would be positioned in close proximity, and an antiparallel arrangement in a pointsymmetrical fashion, which means that the two p15 domains would be positioned on opposite sides of the homodimer. A distance between the centre of gravity of the p5 Is/pS5 d sub-assembly and the piS/pIS sub-assembly was calculated using the parallel axes theorem. Here the assumption was made that there is no drastic difference between the spatial arrangement of the p5 Is/p5 1d sub-assembly in the heteroand homodimers. A similar value of d51,5Id-15/15 = 5.1 i
1136
0.9 nm was obtained for both the parallel and the antiparallel arrangements (Table Ila). This finding is an independent verification of the corresponding parameter in the heterodimer, the distance d5s515d -15, which was 5.1 ±fi 1.2 nm. Even if data analysis showed that the antiparallel arrangement is more likely, it was not possible to decide which of the two arrangements resides in the homodimer.
Discussion RT is one of the major targets in drug design against AIDS (Jacobo-Molina and Arnold, 1991; Mitsuya et al., 1990; Merluzzi et al., 1990; Pauwels et al., 1990). A better knowledge of the structure and function of this enzyme is essential in the search for suitable drugs. Many models have been proposed in the literature on the basis of speculation and suggestions. Our neutron small angle study is the first approach which provides structural information about RT on the basis of experimental data. A crystallographic approach would have been preferable due to higher resolution, but this approach will have to await the availability of suitable crystals. In contrast to the crystallographic approach, the data interpretation of neutron small angle scattering is less straightforward. Detailed conclusions about the shape would require sophisticated fitting procedures which, in general, do not yield unique results. We avoided this by confining the conclusions about the shapes of the components to ellipsoids, since their half axes can be calculated directly from the experimentally obtained scattering parameters. The strength of neutron small angle scattering is the possibility of gettting information about the distances of the components within a molecule. This
Quaternary structure of HIV-1 reverse transcriptase
information is obtained by solution studies without destroying the molecules. According to the capacity of neutron solution scattering it is the aim of this study to determine the spatial arrangement of the domains of RT rather than their shape. The distance determination between two components is possible if the radii of gyration of these components and their assembly are known. The measurement of the radii of gyration of the components in situ requires selective visualization of the components. This is achieved by isotopic labelling of the single components with deuterium. We succeeded in preparing p5 l/p66 HIV-l RT from a protonated p66 subunit and a deuterated p51s subunit using a novel reconstitution procedure (Figure 2, Le Grice et al., 1991; Howard et al., 1991). Subjection of the hybrid H-D-RT to contrast variation with D20 allowed the determination of the radii of gyration of the p66 and p5l, subunits. The radius of gyration of the p15, RNase H-containing domain was derived from molecular modelling procedures using the E. coli RNase H as a basis (Hendrickson, personal communication, see Results). Using the six radii of gyration of the components p15, p51s and p5ld and the assemblies p51s/p66 (the complete heterodimer), p5ld/P15 (the p66 subunit) and the subassembly, p51d/p51,, the centre-to-centre distances between the components were calculated by the parallel axes theorem. Here the assumption was made that there is no great structural difference between the p51s subunit and the p51d domain in situ, justifying the use of the same radius of gyration for both components. We are aware that minor structural differences between the p5ls subunit and the p51d domain must exist, since both components reside within the RT in a different structural context. However, a drastic conformational change which could affect our conclusions on the RT model is rather unlikely, since the expected structural changes are a consequence of a proteolytic excision of one of the p15 components of the homodimeric p66. The relative error of the distances is 10% . The distance between the p66 and p51s subunits is best determined and verified by an independent approach according to Stuhrmann and Miller (1978). The distance between the p15 domain and the sub-assembly, p51d/p51s, shows the least accuracy. This is due to the small influence of the low molecular weight component, p15, on the radius of gyration of the complex. However, this distance is confirmed by an independent distance determination of these components in the p55/p66 homodimer. Despite this, it could not be decided whether the two p66 subunits are aligned parallel or antiparallel. Therefore, the p5ld domain and the p5ls subunit are displayed in the model in Figure 6 as symmetrical bodies. The centres of gravity of the components are located at the corners of the distance triangle formed by the centre-tocentre distances of the components. The orientation of the components was constrained by the overall shape parameters of the heterodimer. Figure 6 shows that RT is a flat molecule with axes of 10 nm, 9 nm and 3 nm. The two p5i components are positioned side by side, attached at their flat sides. The RNase H domain is positioned at the far end of the ellipsoidal RT, most probably having contacts to both p5l components. Figure 6 shows the model of RT in the complex with its template fitting the nucleic acid template in A-form. This model has to be considered as a tentative model with respect to the position of the nucleic acid. The template is docked -
to RT between the two subunits p66 and p5 1, as proposed previously (Jacobo-Molina and Arnold, 1991; Arnold and Arnold, 1991). This position was proposed, taking into account the following information: (i) the functionally active form of HIV RT is the dimeric form (Restle et al., 1990) although the p66 subunit alone contains all functional activities (Cheng et al., 1991; Wu et al., 1991), (ii) there is only one template binding site on the heterodimeric RT (Painter et al., 1990) and (iii) the two subunits p5Is and p66 interact via a hydrophobic region (Becerra et al., 1990; Painter et al., 1990), which might serve as the template binding site. The orientation of RT shown in Figure 6 is based on our knowledge that the RNA template in a DNA -RNA hybrid is cut upstream of the polymerization site (Schatz et al., 1990; Furfine and Reardon, 1991). The number of bases between the RNA cutting site and the polymerization site varies between 7 and 18 bases (Oyama et al., 1989; Schatz et al., 1990; Furfine and Reardon, 1991; Wohrl and Molling, 1990). We used the greatest value of 16-18 bases in order to map the polymerization site in our model.
Conclusion Our neutron scattering studies on selectively deuterated p5I/p66 HIV-1 RT provide information about the spatial arrangement of the components of RT, i.e. the subunit p51s and the domains p51d and p15, in solution. The model derived is considered as a working model, which can serve as a basis for designing new experiments concerning RT structure and function e.g. (i) the role of the p51 subunit in comparison to the p66 subunit, which carries all functional activities of RT, (ii) the conformational change induced in the template following RT binding, (iii) the portion of the template which interacts with RT. Answers to these questions might also be significant for the design of a nucleic acid fragment suitable for co-crystallization of RT with its template.
Materials and methods Preparation of RT Heterodimeric HIV-1 RT was prepared by metal chelate affinity chromatography from E coli strain M 15:: pDMI. 1:: pRT6H-PROT, which directs expression of p66 RT and HIV- 1 protease (Le Grice and GruningerLeitch, 1990). Enzyme thus prepared bears a polyhistidine extension on the C-terminus of the 66 kDa subunit. Reconstituted, selectively deuterated HIV-1 RT was prepared according to published procedures (Le Grice et al., 1991), with the exception that (i) bacteria expressing His-pSI (Schatz et al., 1990) were grown and induced in deuterated algal hydrolysate and (ii) DEAE-Sephacel and S-Sepharose chromatography followed
Ni2+-NTA-Sepharose purification. The protocol for preparation of this enzyme, D-[HisIpSl/Hp66, is outlined schematically in Figure 2a. Homodimeric p66/p66 R was prepared from the strain M 15:: pDM1. 1:: p6HRT as previously described (Le Grice and Gruninger-Leitch, 1990). All enzyme preparations were judged by SDS electrophoresis to be at least 95% homogeneous. Growth and induction of the recombinant E. coli strains was as previously described (Le Grice et al., 1991). In the experiments reported here and related studies (Le Grice et al., 1991; Howard et al., 1991), p5l/p66 prepared by reconstitution retained full polymerase and RNase H activities. Furthermore, tRNALYS.3 binding to the reconstituted, selectively deuterated enzyme was uanffected (J.-L.Darlix, unpublished). The concentration of each protein was determined using an extinction coefficient e1% of 18 for both the homodimer and heterodimer. Data collection The small-angle neutron scattering measurements were performed on the DII small-angle diffractometer of the Institut Laue-Langevin (Grenoble,
1137
H.Lederer et al. France). The samples were dialysed against buffers containing 0.05 M or 0.5 M NaCl, pH 7.9, 10 mM 13-mercaptoehanol in H20, D20 or mixtures of both. During beam exposure the samples were equilibrated to 8°C. The scattering data were processed and corrected as described by Heumann et al. (1988). Concentration effects were eliminated by linearly extrapolating the scattering curves, which were measured at two different concentrations, usually 6 mg/mi and 3 mg/ml, to concentration zero. From the scattering curves the distance distribution functions p(r) were calculated [with a Fortran program of Glatter (1977)], from which the maximum dimensions of the particles were obtained. Forward scattered intensities and radii of gyration were determined from the area and second moment of the p(r) functions, as well as from Guinier plots of ln[I(Q)] versus Q2, where Q is the momentum transfer,
Q = (4) X sin e x where 20 is the full scattering angle. The half axes (a, b, and c) of a scattering equivalent ellipsoid were calculated from the radius of gyration RG and the maximum dimension dmax = 2c using the following equations: R 2 = a2 + h2 + ((1) and
V= 43 r X a x b x c
(2)
The volume (V) of a protein component was calculated from its molecular weight assuming a partial specific volume of v = 0.74 cm3/g (Lehninger, 1975). Distance determination Parallel axes theorem. The centre-to-centre distance d, 2 of two components with the radii of gyration RI and R2 was determined from the radius of gyration R112 of the complexed particle according to the parallel axes theorem of mechanics applied to scattering: Rj,2 =
zi
x
RI
+Z2 z2 -
x Z2 x
dl2
with RC = 10.93 nm, the radius of gyration at infinite contrast, and the coefficients a = -2.58 x 10-4 and 3 = 1.31 x 10-7 nm-2. The quality of the fit is indicated by the correlation coefficient of 0.999. Using the coefficient ( the distance between the components H-p66 and D-pSI is obtained according to the equation
dsl,
- 66 = (
AQi
)
+
( Q2
2
(5)
)
AQ and AQ2 are the contrasts at which the subunits H-p66 and D-p51 matched, respectively. The centre-to-centre distance obtained is
d5l,566
are
= 3.28 + 0.3 nm.
Extrapolation of the plot to the contrast matching of deuterated p51 in a buffer containing the hypothetical D20 concentraton of 118% D20, yielded the radius of gyration, R66 of 3.14 + 0.07 nm. The mass law. Within the p66 subunit, the distance d66-51d between the centres of the component p66 and the p5ld is correlated with the known distance, d66-15, between p66 and p15 according to their molecular weights
M51
and
M15
as
follows:
d66-51d x
M51
=
d66-15
x
M15
(6)
Estimation of the distance between the centres of the pSJd domain and the pSI, subunit, and of the radius of gyration Rs5d/s,5. The centre of the p5 1d domain lies on the sphere of radius d51d-66 = 1.1 0.1 nm centred around the p66 subunit. This distance is calculated according to the mass law (6). An upper and lower limit of this distance is therefore, d515_51d = d5is _66 d5ld-66 = 3.3 i 1.2 nm. The upper limit, (d5l 51d))max .. is 4.5 nm. This can be further restricted to a value of 3.9 nm using the parallel axes theorem (3) by taking into account that an upper limit of the radius of gyration of the p5ld/p5l, sub-assembly is the radius of gyration of the complete heterodimer, R66/51s = 3.27 + 0.04 nm. It then follows that: 2.1 nm
