,I. Mol. Biol. (1992) 227, 719-737

Structural

Requirements

for Viroid Processing by RNase Tl

G. Steger’, T. Baumstark’, M. Miirchen’, M. Tabler2p3, M. Tsagris2*3, H. L. Stinger’ and D. Riesner’ ‘Institut fiir Physikalische Biologic Heinrich-HeineUniversitdt Diisseldorf Universittitsstr. 1, D-4000 Diisseldorf, F.R.G. 2Max-Plamk Institut fiir Biochemie D-8033 Martinsried, F.R.G. 3Foundation for Research and Technology-Hellas Institute of Molecular Biology & Biotechnology Heraklion 711 10, Crete, Greece (Received 1 April

1992; accepted 5 June 1992)

Viroids are replicated via a rolling circle-like mechanism in which ( + ) strand oligomeric intermediates have to be cleaved enzymatically to unit-length molecules followed by ligation to mature circles. A transcript of potato spindle tuber viroid, which is still infectious, consists of a monomeric molecule with only 22 additional nucleotides, thus doubling part of the central conserved region of viroids. It was shown that this transcript can be cleaved and ligated in vitro to circles by RNase Tl. To elucidate the site and mechanism of processing, 16 different site-specific mutants of this longer-than-unit-length transcript were constructed and analyzed by in vitro processing with RNase Tl, infectivity studies, temperature-gradient gel electrophoresia, and structure calculations. The wild-type sequence and several mutated transcripts are able to adopt a particular secondary structure which is the prerequisite for enzymatic cleavage and ligation by RNase Tl. This “processing structure” exposes both potential cleavage sites in the nearest spatial neighborhood, thus favoring the subsequent ligation to circles. Those mutated sequences for which the formation of the processing structure is impossible or thermodynamically highly unfavored are not processed. The results demonstrate that the particular structural features of viroids enable them to be cleaved and ligated by one and the same enzyme, RNase Tl. The in vitro mechanism may serve as a mechanistic model for cellular processing of viroids. Keywords: viroid;

RNA secondary structure; RNA processing; RNase Tl; temperature-gradient gel electrophoreais

reviews, see Diener, 1979, 1987; Sanger, 1982; Riesner & Gross, 1985; Riesner & Steger, 1990). The current model of viroid replication assumes a rolling circle mechanism (for reviews, see Gross & Riesner, 1980; Siinger, 1987; see Fig. 1). The circular (by definition ( + ) strand) viroid is transcribed into oligomeric ( - ) strand RNA. The ( - ) strand acts as a template for the synthesis of an oligomeric (+) strand RNA. Both transcription steps are catalyzed by a host enzyme, the DNA-dependent RNA polymerase II (Miihlbach & Sanger, 1979; Rackwitz et al., 1981; Semancik & Harper, 1984; Schindler & Miihlbach, 1990). The ( + ) strand oligomeric RNA is cleaved enzymatically to unit-length molecules, which are then ligated to the mature viroid circles. Self-cleavage and self-ligation could not be verified

1. Introduction Viroids are plant pathogens distinguished from viruses by the absence of a protein coat and by their small size. They are circular, single-stranded RNA molecules consisting of a few hundred nucleotides, the smallest being about 240 and the largest about 600 nucleotides. Obviously viroids can only possess a very limited coding capacity, and there is no experimental evidence for a viroid-encoded tranalation product. Thus, one has to assume that viroid replication and pathogenesis completely depend on the enzyme systems of the host. Their “genetic” information must be the RNA structure, the ability to undergo particular structural transitions, and the capability of interacting with host cell factors (for 0022%2836/92/190719-19

$08.00/O

719

0 1992 Academic: Press Limited

G. Steger et al.

Figure 1. Replication cycle of potato spindle tuber viroid (PSTVd). The infecting circular molecule is transcribed into oligomeric ( - ) strands (steps 1 and 2) followed by synthesis of (+) strand oligomers (step 3). The ( + ) strands are cleaved enzymatically into monomeric units (step 4) and ligated enzymatically to mature ( + ) circles (step 5). The Figure is reproduced from Branch & Robertson (1984).

in spite of several attempts (Robertson et al., 1985; Tsagris et al., 1987). This is in contrast to the processing of avocado sunblotch viroid: unit-length molecules are formed by self-cleavage of oligomers; no enzyme has to participate during this processing step (Hutchins et al., 1986; Forster & Symons, 1987; Symons, 1990). One has to assume that the accuracy of cutting and ligation of oligomeric replication intermediates is a consequence of a well defined secondary structure of the sites. It should be emphasized that viroids, as molecular parasites, do not provide their own highly specific processing factors, and that the host cell processing factors are not adapted to viroid replication. Thus, in clear contrast to host RNA processing, the secondary structure of viroids and their replication intermediates have to play a dominant functional role. The structure and structural transitions of viroids are known in some detail. Under native conditions, viroids form a rod-like structure, which may be described as a serial arrangement of short helices and small internal loops. The secondary structure of the potato spindle tuber viroid (PSTVdt) is shown in Figure Z(a). During thermal denaturation, viroids undergo several structural transitions from the rodlike structure to the single-stranded circle without any intramolecular base-pairing (Hence et al., 1979; Riesner et al., 1979). In a highly co-operative main transition, all base-pairs of the native structure are disrupted and particularly stable hairpins are newly formed (hairpin I, II, III). This transition may be 7 Abbreviations used: PSTVd, potato spindle tuber viroid; UCCR, upper central conserved region; nt, nucleotide; LCCR, lower central conserved region; TGGE, temperature-gradient gel electrophoresis; HP, hairpin.

viewed as a switch from an extended to a branched structure with a marked loss of base-pairing. At higher temperatures the stable hairpins dissociate independently from each other in the order of their individual thermal stabilities. Hairpin I and hairpin II are more conserved between different viroids than the rest of the molecule with regard to position, length and G +C content. Hairpin III was found only in PSTVd and is therefore of marginal interest. Recently, the functional relevance of a hairpin II-containing structure of viroids was studied by site-directed mutagenesis, t,hermodynamic methods and infectivity tests (Loss et al., 1991). It had been pointed out that hairpin II is formed not only during thermal denaturation of circular viroids, but also, and probably biologically more important. as part of a metastable structure during the synthesis of viroid replication intermediates. Mutations in the segments that form hairpin II and infectivity tests with the mutated viroid cDNA showed that the core region of the stem of hairpin II is critical for viroid replication. Taking into account literat,ure data on binding sites for transcription factors, these investigations led to the hypothesis of hairpin II fumtioning as a binding site for host cell transcription factors. Several reports in the literature indicated that the region containing hairpin I may be involved in the processing of oligomeric replication intermediates (Tabler & Siinger, 1984; Visvader et al.. 1985: Diener, 1986; Steger et aZ., 1986; Hecker et nl., 1988). Hairpin I is formed in a region of the molecule t,hat shows a strong sequence homology among all viroids of the PSTVd-class. In Figure Z(a) this segment is designated the upper central conserved region (UCCR). In the case of PSTVd, it was reported recentI!, that RNase Tl is able to catalyze in vitro both the cutting and the ligation reaction without requirement of any other protein (Tabler & Tsagris, 1990: Tsagris et nE., 1991). As a working hypothesis for the processing reaction, the involvement of similar enzymes can be envisaged within the plant. cell. Furthermore, since PSTVd is representative in many aspects of a large class of viroids. the processabil&y by RNase Tl might also hold for other viroids. For their study, Tsagris et nl. (1991) used a longer-than-unit-length linear t#ranscript (nt, 86 t,o 359/l to 106 with an additional 12 nt at the 5’ end and an additional 13 nt at the 3’ end). Thus. the transcript contains the central part of the UCCK twice, at the 5’ and at the 3’ ends (Fig. 2(h)). FOUJ of the E&vector nucleotides adjacent’ to the PSTVd-specific sequence are identical to the PSTVd sequence G80 to U83, so that only the CX4 is absent from a 5 nt longer PSTVd-specific stretch at the 5’ end of the transcript. The position G80 was determined as the site of cutting and religation of a (+ ) strand RNA transcript by RNase Tl. Since this transcript contained several G residues at which n priori cutting and ligation would lead to exact PSTVd circles. but the complete reaction was

5’

289

“’ AJcGGAG-Gcu”c lllll - YGCC c ”

69

11111 CGAAG.,,.

/

.a

9..“.=.‘gc”“*cGGA”ccc~

Lb””

\

117

263

AGCG*A’CUGG&I I I I I I II ” GUCG-, a-e ,cuUCGC

CGGGGAAACCUGG

Figure 2. Structures of PSTVd and sequence of the longer-than-unit-length transcript Ha106. (a) Optimal secondary structure of PSTVd (KF440-2 lethal) at 50°C. Palindromic sequence elements I and I’, II and II’, III and III’ (see lines), respectively, are able to form hairpin helices. Nucleotides, conserved in PSTVd, citrus exocortis viroid, chrysanthemum stunt viroid, tomato apical stunt viroid, tomato planta macho viroid, and grapevine viroid, all belonging to the PSTVd class, are given in bold face (upper and lower central conserved region, UCCR and LCCR, respectively). The UCCR contains a GC palindrome, GCP. Nucleotide numbering is given according to Gross et uZ. (1978). (b) Partial sequence of the transcript Ha106 showing the duplication of the PSTVd-sequence from nucleotide 85 to 106 as well as the vector-derived sequences at 5’ and 3’ ends (lower case letters). Extensions of the transcript to the left and to the right are identical to PSTVd.

(b)

722

G. Steger et al. scripts as model oligomers (Steger et al., 1986; Hecker et al., 1988). It was suggested in several reports that formation of the tri-helical structure is a prerequisite for processing (Diener, 1986; Steger et aE., 1986; Hecker et al., 1988). In the present study four types of specific mutations were introduced in the UCCR and one in the lower central conserved region (LCCR), which is opposite to the UCCR in the native rod-like st)ructure. By thermodynamic calculations and temperature-gradient gel electrophoresis the effect, of the mutations on the secondary structure was determined. These data were compared with the in vitro processability of the different linear transcripts. A model for the struct,ure of the processing site could be derived, which is clearly different from our earlier expectation and the literature reports.

2. Materials and Methods (a) Transcripts

Figure 3. Central part of the optimal secondary structure of the transcript Ha106 at 50°C (cf. Fig. 2(b)). For I, I’ and GCP see Fig. 2(a)). The numbers 1, 2, 3, 4 and 5 denote sites of mutations in the transcript Ha106. These mutations, G to C or C to G, respectively, were selected in order to change the thermodynamic stability of the depicted secondary structure. The transcript Ha106-C84 contains a C at position 84 in addition to the sequence of Ha106 (see open arrow).

observed only at G80, one has to conclude that the reaction is guided by a particular secondary structure of the transcript. Therefore, the basic question of the present work, is to determine the secondary structure of this longer-than-unit-length transcript as a prerequisite to understand the molecular processing mechanism of viroids. Then, this structure could serve as a model for the structure of natural oligomeric replication intermediates. Studies on the structure of the UCCR in oligomerit replication intermediates were carried out before (Steger et al., 1986; Hecker et aE., 1988) and were actually the starting point of the present investigation. Since the UCCR and the neighboring sequences on both sides are extensively self-complementary, two UCCRs, either from two different or from two different unit-length molecules segments in one and the same oligomer, are able to form a very stable tri-helical region of 28 base-pairs interrupted by two small internal loops (see Fig. 3). Applying thermodynamic calculations, optical and temperature-gradient gel melting curves electrophoresis, the presence of this tri-helical structure has been verified in vitro with dimeric tran-

of pHalO6 and its derivatives

Construction of the plasmid pHalO6 has been described in detail before (Tabler & Sanger, 1985: Tsagris et al.. 1991). The cDNA of PSTVd KF440-2 (SchnGlzer et al.. 1985) was inserted into the HindITI/&oRI sites of the vector pGEM I (Promega) and transcribed in vitro with SP6 polymerase after EcoRI-linearization, RKA synthesis and purification followed published protocols (Melton et al., 1984; ijfverstedt et al.. 1984; Tabler t SPnger, 1985). The mutations introduced by site-directed mutagenesis (Kramer et al., 1984) are given in Fig. 3 in which the UCCR and LCCR are presented in the tri-helical structure. All transcripts studied in this work are listed in Table 1. The numbers behind HalOfi- designate the sites of G to C or C to G exchange, respectively, as depicted in Fig. 3. All except 4 of the transcripts contain 406 nucleotides (nt), consisting of 12 vector-derived nt at bhe %-terminal region, the PSTVd-specific sequence starting with nt 85. one complete PSTVd-unit, nt 86 to nt 106 (as a repeat). and 13 vector-derived nt at the 3’ terminus. The total sequence of the transcripts may be seen from Fig. 2(b). Three of the transcripts (Ha106-14. -15. -1245 and -12345) are transcribed in vitro from a cDNA cloned into pGEM 2 with T7 polymerase after EcoRI-linearization. Thus, they contain 16 vector-derived nt, at, the 5’ end (5’ GGGAGACCGGAAGCUU 3’ instead of 5’ GAAUACAAGCUU 3’), giving a total number of 410 nt. In addition to the large series of mutations, one construct (pHa106X84) was made by inserting a (’ between the $-terminal vector-derived sequence and nt 85 of the PSTVd sequence yielding a 407 nt transcript (Tabler el al.. 1992). Due to the insertion of C84 and the 4 adjacent nucleotides of the 5’-terminal vector sequence. which are identical to GSO. C81, U82 and U83 of PSTVd. the PSTVd-specific sequence starts at GSO (instead of A85). continues throughout one unit and the repeat from nt 80 to nt 106 yielding a 386 nt PSTVd-specific transcript. The sequences of all mutants were confirmed by DNA sequencing. (b)

Calculation

of secondary

structures

The theoretical analysis of the secondary structures and the structural transitions of all transcripts was carried out using an algorithm that was described in detail (Schmitz

Structural

Requirements for Viroid Processing

& Steger, 1992). The main features of the algorithm will be outlined together with the results. All calculations were carried out on a VAXstation II (Digital Equipment Corporation). (D) Temperature-gradient

723

@lS% (w/v) bisacrylamide, @l% (v/v) TEMED, @2 x TBE (17.8 m&f-Tris, 17.8 mnn-boric acid, @4 mMEDTA) and @07% (w/v) ammonium peroxodisulfate for starting the polymerization. Silver staining of gels was performed according to Schumacher et aE. (1986).

gel electrophoresis

Thermal denaturation profiles were determined by temperature-gradient gel electrophoresis (TGGE) (Rosenba,um k Riesner, 1987; for reviews, see Riesner et al., 1989, 1991). As outlined before (Hecker et al., 1988) the method is particularly suitable for the analysis of coexisting secondary structures. A horizontal slab gel (185 mm X 175 mm X 1 mm) on a film support is in thermal contact with, but electrically insulated against, a thermostatting metal-plate. A linear temperature gradient is established perpendicular to the direction of electrophoresis using 2 thermostatting baths, each connected to one of the opposing edges of the plate. For this work, a commercially available instrument was used (DIAGEN-TGGE System, DIAGEN, 4010 Hilden, FRG). The temperature gradient was between 20°C and 57°C. The RNA sample (200 to 500 ng in 800 ~1) was pipetted into the broad sample slot (150 mm x 5 mm); for size reference natural PSTVd isolate (10 to 30 ng in 40 ~1) was applied into 2 small slots (5 mm x 5 mm) at both sides of the broad slot. Gels contained 5% (w/v) polyacrylamide,

(d) Ionic strength dependence All calculations are performed for 1 m-ionic strength; TGGE was carried out in @2x TBE (equaling about 2.9 IIIM ionic strength). An increase in the transition temperature of viroids of 13.5 deg.C per order of magnitude in ionic strength was reported earlier (Langowski et al., 1978; Seger et al., 1984). Thus, a difference of 33.4 deg. C between calculated and measured transitiontemperatures has to be taken into account.

3. Results and Discussion (a) Position of mutations in longer-than-unit-Zen& infectious PST Vd speci$c RNA transcripts In order to clarify the involvement of the hypothetical structure of the processing site and to receive more information about the precise location of cutting and ligation, specific mutations were

Table 1 Correlation

of in vitro

processability,

structure

in TGGE, and AGO-calculation of transcripts Ha106 and its

mutants -AGO (50°C) [kJ/mol]

structure In

Transcript

vitro

processability

predominant in TGGE

TH

ExL

ExM

ExR

Ha106 Ha106-3 Ha106-13

+ +

TH, ExM TH ExM

364 364 347

304 302 302

321 318 321

335 332 340

Ha106-4 Ha10&14* Ha106-134

+ +

ExM ExR ExM

347 329 329

317 320 314

337 332 335

332 340 343

Ha106-25 Ha106-235

+ -

ExM ExR

329 329

304 302

321 318

335 332

Ha106-12345* Ha106-1245*

+ -

TH, ExM TH

364 364

322 320

335 332

343 340

Ha106-4 Ha106-24 Ha106-5 Ha106-15*

+ + -

ExM TH, ExR ExM TH

347 364 347 364

317 317 304 308

335 335 321 318

332 332 335 343

Ha106-45 Ha106-345

++ +

ExM ExR

329 329

317 315

335 332

332 335

Ha106-123 HalOt-

++ +

ExM ExR

329 329

302 300

321 318

340 343

Ha106-C84

+

TH, ExL’

356

326

321

334

The numbers 1, 2, 3, 4 and 5 behind Ha106- denote the presence of G to C or C to G exchanges at sites as specified in Fig. 3. The vector-derived sequence at the 5’ terminus of transcripts marked with an asterisk deviates from those of all other transcripts (see Materials and Methods). In vitro processability: Circular molecules of 358 nt are produced by RNase Tl in significant amounts ( + ). in higher amounts ( + + ), or are not detectable (-). In the case of Ha106-C84 the circular product has the correct size of 359 nt. Structure predominant in TGGE: If 2 structures are given, pretreatment in low ionic strength favors the extended structure, pretreatment in high ionic strength the tri-helical structure. Nomenclature of structures and transition curves according to Figs 3 and 8 and Figs 4 to 6, respectively. TH: curve 5 (tri-helical structure); ExM: curve 2 (extended middle structure); ExR: curve 3 (extended right struct#ure); ExL: curve 1 (extended left structure). The superscript 1 in the case of Ha106-C84 emphasizes that also ExM and ExR were visible in low amounts, and even in higher amounts after pretreatment in low ionic strength (cf. Fig. 6). The AC values refer to the secondary structures shown in Fig. 8. If the transcripts Ha-14, -12345, -1245 and -15 had the same r-terminal sequence as the other transcripts, their AC values for the ExL-structure would be 8 kJ/mol more positive.

724

G. Steger et al.

introduced into the plasmid pHalO6. The transcript Ha106 is infectious if inoculated on tomato plants and active as a substrate in the processing reaction with RN&se Tl. Altogether, the wild-type and 15 mutated PSTVd-specific RNA transcripts were synthesized. The consequences for in vitro processability as well as for the secondary structure of the transcripts of the mutated plasmids were analyzed. The starting point for the mutations was the transcript of plasmid pHa106. The most stable secondary structure of the transcript without mutations contains a tri-helical region. This structure, as described in earlier reports, as well as the sites of mutations are depicted in Figure 3. The sites of mutations in Ha106 were selected to affect the tri-helical region. Single-site mutations should destabilize the tri-helical region while particular pairs of mutations are compensatory; i.e. replace G . C pairs by C * G pairs. We selected the nucleotides C91 and G98 (the numbering of mature PSTVd according to Gross et al., 1978), which were either mutated within the 5’-terminal repeat (mutation site 1 and 2, respectively; see Fig. 3) or within the 3’-terminal repeat (site 4 and 5, respectively). Any of these mutations disturbs the tri-helical structure. However, in transcripts where either the sites 1 and 5 or the sites 2 and 4 have been changed simultaneously, the mutations are compensatory; thus restoring the stability of the tri-helical region. An additional mutation was introduced in the LCCR at position G269, which forms a base-pair to C91 (mutation site 1) in the native rod-like structure, therefore compensating the effect of mutation 1. Because nucleotide G98 is unpaired in the native structure, no compensatory change t,o this residue was necessary. During the course of this work. it was found that the products of the RNase Tl cutting-ligation reaction were circular molecules of 358 nucleotides missing C84 (Tsagris et al., 1991). These circles were not infectious. Therefore, a different type of mutant (pHa106C84)) was constructed with C84 inserted at the 5’ site of the linear transcript (see Fig. 3). This transcript resulted in infectious circles of 359 nucleotides after the RNase Tl treatment (Tabler et al., 1992). (b) Structural analysis by TGGE TGGE of the transcripts showed several different yet recurring types of transition curves. Throughout the description of this and other TGGEs, a uniform scheme of numbering the transition curves is applied, i.e. the same type of transition curve is always designated with the same number. The molecular shapes and the conformations of the transcripts may be identified due to the characteristic mobilities and transition temperatures. (i) TGGE of the wild-type transcript Hal06 TGGE of the wild-type transcript Ha106 is depicted in Figure 4(a). Five different transition curves can be detected. Since all bands comigrate at

the highest temperature, i.e. under complete denaturation (compare linear PSTVd in the marker slot; for band 8’ see Figure legend), all curves have to belong to an RNA molecule of unique size. This demonstrates that the different transition curves represent structures of one and the same molecule. which co-exist under the solution conditions of the electrophoresis. Although we will give a quantitative interpretation of the transition curves later. a few qualitative arguments to interpret transition curves on the basis of particular secondary structures should be outlined here. The most prominent band (5) is retarded at. a temperature of 30°C. At higher temperature it’ migrates as slowly as the fully denatured circle of natural PSTVd (compare marker slot at high temperature). At a remarkably high temperature of 47 “C the corresponding structure denat,ures within a discontinuous transition. The only struct,ural element that may be stable up to that t,emperature is the tri-helical structure. Therefore. one has to attribute the tri-helical structure (Fig. 3) to band (5). Band (2) migrates the fastest and shows two transitions at intermediate temperatures. The high mobility of this band suggests a rod-like st’rmture without, any bulky bifurcations. Bands (6) and (7) migrate extremely slowly and finally denature at the same t,emperature as band (5). Their low mobilities point to bimolecular complexes. Because the dissociation temperature is characteristic for thta trihelical region, one should expect that two monomeric transcripts form one or two tri-helical regions in an intermolecular mode. (ii) Shift in the relutivr cowentrations qf co-existing dructurcs by pretreatment in different ionic strengths The relative concentrations of‘ different Wexisting structures were shifted drastically by incubation of the sample in low or high ionic strength prior to TGGE. This is shown in Figure 4(b) and (c). Heating of the sample in high ionic strength (see legend to Fig. 4(b)) at a t’emperature just below the transit’ion at 47°C and cooling to the lowest temperature of TGGE, shifts the relative poncentra.tions remarkably. Transition curve (2) vanished completely, and t,he intensity of curve (4), which was barely detectable in the TGGE without pretreatment, is increased markedly. Curve (4) merges into curve (5) at an elevated t’emperaturr but below the transition at, 47 “C. For some mutated transcripts, incubation in high ionic strength favored the formation of bimolecular complexes. This effect will not be followed further since bimolecular complexes were not the main interest, of this work. As an alternative pretreatment, complete denaturation and renaturation of the sample was carried out in low ionic strength (see legend of Fig. 4(c)). The concentrations of the structures were shifted into the opposite direction (see Fig. 4(c)) as compared to the incubation in high ionic strength (Fig. 4(b)). Quite generally, one might expect that high ionic

725

Structural Requirements for Viroid Processing

V (b)

20 oc

T -

57OC

Figure 4. TGGE of the transcript of Ha106. For conditions of TGGE, see Materials and Methods. Marker slots (left and right side of the broad sample slot) contain natural circular and linear PSTVd. The individual transition curves are designa,ted with numbers 1 to 8; the same numbers refer to similar transition curves throughout all TGGEs. (a) Transcript without any further pretreatment (see Materials and Methods). (b) Transcript after heating and cooling in high ionic strength prior to electrophoresis. Conditions of pretreatment: 500 mM-NaCI, 4 M-urea, 91 mM-EDTA, 1 m&I-NaCacodylate (pH 69); 5 min at 75°C; renaturation with 025 deg.C/min. (c) Transcript after heating and cooling in low ionic strength prior to electrophoresis. Conditions of pretreatment: 10 m&r-NaCl, 0.1 mM-EDTA, 1 mMKaCacodylate (pH 6.9); 5 min at 95°C; renaturation with about 95 deg.C/min. strength favors bulky structures, e.g. branched structures with the tri-helical region, and that low ionic strength induces more extended structures such as the native, rod-like structure without bifurcations. (iii) TGGE of the mutants of the wild-type

transcript

Ha106 The mutations introduced into plasmid pHalO6 drastically influence the transition curves of the corresponding transcripts. In Figure 5, TGGE of three different mutants is depicted. These examples were selected because their transition curves belong

to different prototypes. The mutated transcript Ha106-3 (Fig. 5(a)) exhibits a transition very similar to curve (5) of the wild-type (Fig. 4(a)). The transition curve of Ha106-25 (Fig. 5(c)) is nearly identical to curve (2) of the wild-type. The dominant transition of Ha106-235 (curve (3) in Fig. 5(b)) is similar to curve (2) in that it starts from a fast moving conformation at low temperature and is less retarded than curve (5). It may, however, be differentiated clearly from curve (2), because the retardation occurs in one step in curve (3) but in two steps in curve (2). All 15 mutated transcripts exhibited one of the three types of curves: (2), (3) or (5).

726

G. Steger et al.

(a) cPSTVd

PS TVd V

(b) cPSTVd

PS TVd

IPSTVd

(cl cPS TVd

PS TVd

20 OC

T

l

F‘igure 5. TG( GE of tram script (a) Ha106-3, (b) Ha106-235, and (c) Ha106-25. terr ninola 1gy of tr ,anscripts ser: Table 1 and Fig. 3. Otherwise see Fig. 4. Following the qualitative arguments as outlined above, one would expect that the mutated transcripts may assume either the tri-helical structure (curve (5)) or two different extended, rod-like structures (curves (2) or (3)). (iv) TGGE of transcript Halo&C84 When the PSTVd-specific sequence was extended at the 5’ region by C84 (Ha106-C84), two clear observations were made in TGGE of the corresponding transcript. First, a new type of curve ((I) in Fig. 6) appeared as the dominant band; evidently, it belongs to a fast moving structure and exhibits one transition at low temperatures and a second discontinuous transition at intermediate temperature. This structure is similar to type (2) and (3) and should be of a rod-like type. Second, this transcript is able to assume all types of secondary structures (1) to (5) as co-existing struc-

57T all witho ut pretreatme :nt. For

tures. While the TGGE in Figure 6 shows only the structures (l), (2) and (3) arising from low ionic strength pretreatment of the sample, TGGE after high ionic strength pretreatment reveals the secondary structures (4) and (5), which are characterized by the high thermal stability of the tri-helical element (data not shown). Bimolecular complexes were also found after high ionic strength pretreatment. (c) Attribution

of secondary structures curves in TGGE

to transition

(i) General rules

In order to attribute secondary structures to transition curves from TGGE, thermodynamic features as well as gel-electrophoretic mobilities have to be taken into account. Whereas the transition temperatures may be calculated quite accur-

727

Structural Requirements for Viroid Processing

IPSTVd

8

T _______,

20°C Figure Fig. 4.

6. TGGE of transcript

of Ha106-C84 after pretreatment

of gel-electrophoretic ately, the interpretation mobilities has to rely more on qualitative arguments. (1) Branched structures migrate slower than extended structures. This effect is known from the denaturation of double-stranded nucleic acid, which leads to drastic retardation as long as the denaturation is incomplete (Steger et al., 1987). The effect has been described also with dimeric transcripts of PSTVd (Hecker et al., 1988). (2) Structures with large loops migrate extremely slowly. The low mobility of denatured circular viroids and the lower mobility of plasmids in the form of a relaxed circle as compared to supercoils are examples of this tendency. (3) Because of their higher molecular weight and their usually high degree of bifurcations, bimolecular complexes migrate much slower than the corresponding uncomplexed molecules. (ii) IdentiJication of the b-i-helical structure The most obvious thermodynamic feature of curve (5) is the transition at high temperature (47 “C). From previously published calculations (Steger et al., 1987; Hecker et al., 1988) it could be derived that only the tri-helical region (Fig. 3) is stable up to that temperature. Consequently, the transition at 47°C is a clear indication for the presence of the tri-helical structure. Before denaturation of the tri-helical region but after all other basepairs are dissociated, the secondary structure consists of a large loop similar in size to denatured, covalently closed circular PSTVd. Therefore, the extent of retardation is similar to that shown by circular viroids. Furthermore, TGGE of the transcripts after pretreatment in different ionic

in low ionic strength

60°C (see Fig. 4(c)). Otherwise, see

strengths, supports the interpretation given above. Pretreatment in high ionic strength increases electrostatic shielding of the phosphate backbone and thereby favors bulky or branched structures, and the formation of bimolecular complexes; pretreatment in low ionic strength always shifts the equilibrium to extended structures because the less shielded negative charges try to gain maximum distance from each other. In accordance with this expectation and the attribution of curve (5) to the tri-helical structure, the intensity of curve (5) is highest after pretreatment in high ionic strength (Fig. 4(b)) and lowest in low ionic strength (Fig. 4(c)). Curve (4) (in Fig. 4(b)) coincides with curve (5) after a denaturation step at 37°C. Thus, it exerts the characteristic dissociation of the tri-helical region at 47°C and has to contain that structural element. Because formation of this structure is even more dependent upon renaturation in high ionic strength than that of the structure of curve (5) (compare Fig. 4(b) and 4(a)) one has to assume that the structure of curve (4) is more bulky than that of curve (5). The most obvious assignment of curves (4) and (5) to secondary structures would be the following: the structure of curve (5) is identical to the one in Figure 3; i.e. what originally was called “tri-helical structure” and served as the concept for selecting the mutations. It contains the UCCR and LCCR in the tri-helical conformation and on both sides the rod-like structure of native viroids. The structure of curve (4) may be regarded as being collapsed from the partially denatured state in which the strand forms a large hairpin loop with the tri-helical region as stem. In accordance with this interpretation, the denaturation of the collapsed

728

G. Steger et al.

structure (curve (4)) is less co-operative than that of the tri-helical structure (curve (5)), which exerts a co-operativity characteristic for the unbranched left and right halves of the viroid structure. Since, in contrast to our early expectation, neither structure discussed above was active in the biological processability test (see below), we are not going to interpret their transition curves in more quantitative terms. (iii) Identijkation of extended secondary structures Curves (1 ), (2) and (3) represent structures with high electrophoretic mobility at low temperature (see Figs 4 to 6). Since their mobility is markedly higher compared to that of the tri-helical and the collapsed structure, we assume that the structures of curves (1 ), (2) and (3) are extended structures similar to the native rod-like PSTVd structure. They all show one or two transitions with decreasing mobility at intermediate temperature and one transition with increasing mobility at higher temperature. Since the electrophoretic mobilities of curve (1) to (3) are very similar while the temperatures of the transitions are different, the attribution of the curves to distinct structures could be achieved only by detailed thermodynamic calculations. (d) Thermodynamic

calculations

An algorithm was applied, which is described in detail by Schmitz t Steger (1992); it is an extension of the algorithm of Nussinov and Zuker (Nussinov & Jacobson, 1980; Zuker & Stiegler, 1981) and its modified form from Steger et al. (1984). For each of n discrete temperatures within a chosen temperature interval, the 30 most stable secondary structures were calculated. After purging of multiply occurring structures, a structure ensemble of maximally n x 30 different secondary structures was the structure created. Typically, ensemble contained 200 to 220 different structures calculated at 50°C up to 95°C in intervals of 5 deg. C. For each structure the optical absorbance A,,,, the free the statistical weight energy AGO and exp( - AG’/kT) within the population of all structures were calculated. Thus, a plot of the weighted optical absorbance of all structures versus temperature results in an optical denaturation curve. It may be compared with experimental optical melting curves and TGGE. The probability of the individual nucleotides to be involved in base-pairing is calculated using their statistical weights in the structure ensemble and is represented in so called “probability plots” described below. (i) Base-pair probability plots Base-pair probability plots p(i, j) are an extension of the well known Tinoco-plots (Tinoco et al., 1971) to which the thermodynamic probability of each base-pair is added as the third dimension. The matrix of the combinations of base i and base j forming a base-pair is not restricted to the

400

3 ExR

1

50

100 150 200 250 300 Nucleotide i

350 400

Figure 7. Three-dimensional base-pairing plot p(i. j) of transcript of Ha106 at, 50°C. The base-pairing probability, p, is plotted versus the sequence in 5’ to 3’ direction, i, versus the sequence in 5’ to 3’ direction: j; sequence numbering is from 5’ to 3’; i and j form a plane comparable to a Tinoco-plot (Tinoco et al., 1971) with p arranged

vertically to this pIane. Structures containing the trihelical region (see Fig. 3) were eliminated by excluding the base-pairs of the GC palindrome. GCP, from the matrix of all possible base-pairs. Helices are represented by peaks and their base lines perpendicular to the diagonal. Only helices with a probability above @Ol are displayed. LH, RH: Left, and right half of the secondary structure; they are identical to the most stable structure of circular PSTVd (see Fig. 2(a)). ExR: Terminal helix connects 5’ and 3’ ends of the sequence. ExL, ExM: Terminal helices which are formed from nucleotides of the 3’ end of the sequence. ExR, ExL and ExM are alternative structures. The corresponding secondary structure models are shown in Fig. 8.

secondary structure of lowest free energy, but attributes individual probabilities to all base-pairs of the ensemble of structures. A probability value of 1.0 for a combination of base i and base j, for example expresses the fact that all molecules at, a given temperature contain the base pair i.j. A probability value of @5 for example could indicate that this base-pair is dissociated already to 509/o or that at least one of the bases is involved in another basepair with the remaining probability. Both possibilities may also occur simultaneously. In Figure 7 the p(i,j) plot of the transcript Hal06-C84 is given for 50°C in 1 M ionic strength, which corresponds to native conditions (X 30°C. z 100 mM ionic strength). For a reliable comparison of experiment and theory, however, some restrictions have to be introduced into the algorithm in order to account for the influence of the low ionic strength on the structures. (ii) incorporation of conditions of low ionic strength Several effects have to be considered when theoretical results obtained at 1 M ionic strength are compared with experimental results from TGGE carried out in @2x TBE, equaling about 2.9 MM

Structural Requirements for Viroid Processing

729

Extended rinht structure

( b ) Extended middle structure A **SC