Deletion Analysis of Qβ Replicase

We have analyzed one of the functional domains of Qβ replicase, an RNA-dependent RNA polymerase of RNA coliphage Qβ. Deletion mapping analysis of the carboxyl-terminal region of the β-subunit protein revealed that the terminal 18 amino acid residues (positions 571–588) are dispensable for the replicase reaction. Subsequent deletions up to the Ala-565 residue reduced the RNA polymerizing activity of the replicase in vivo but increased it in vitro. The mutant replicases with enhancedin vitro RNA polymerizing activity were found to have relaxed template specificity for ribosomal RNAs and cellular RNAs as well as Qβ RNA. Deletions beyond the Ile-564 residue abolished both the RNA polymerizing activity and the binding ability to midivariant (MDV)-poly(+) RNA, a derivative of a natural template for Qβ replicase, MDV-1 RNA. These results suggest that the carboxyl-terminal part of the β-subunit participates in RNA recognition of Qβ replicase.

With the development of modern gene technology, it became feasible to analyze the relationships between the structures and functions of gene products. In the case of RNA coliphages, the complete nucleotide sequences of representative phage genomes were determined (1)(2)(3)(4), and analysis of the structure of each gene also became possible. In attempts to clarify the functional domain of the polymerases, several polymerase genes have been extensively characterized.
An RNA-dependent RNA polymerase (RNA replicase) of RNA coliphage Q␤ consists of four subunits. Three of them are host-derived proteins (ribosomal protein S1, and protein elongation factors, EF-Tu and EF-Ts) (5). Only the ␤-subunit, which is composed of 588 amino acid residues, is encoded by phage RNA. The central region of the ␤-subunit is well conserved in RNA coliphages, and this region is thought to contain common structural features for such as the assembly of the subunits and catalysis for RNA synthesis (6). A sequence motif, Gly-Asp-Asp, which has been identified in RNA-dependent RNA polymerases of many viruses and is considered to be the active site for polymerization (7,8), is located in the central part of the ␤-sububit protein (9). Recently, this motif sequence of the Q␤ and polio virus 3D pol proteins was found to take part in the binding of metal ions necessary for RNA polymerization (10,11).
The Q␤ replicase can specifically transcribe in vitro the Q␤ RNA as well as the RNA from closely related phages (5). S1 and another host protein, HF-I (12), are required for the recognition of only Q␤ viral plus-strand RNA as a template. In contrast, Q␤ replicase lacking S1 and HF-I transcribes poly(rC), MDV-1 RNA, and Q␤ minus-strand RNA, as does the holoenzyme of Q␤ replicase (5). RNA replicases from phages MS2, GA, and SP share the host subunits of Q␤ replicase except for HF-I, but they have the different template specificity. Therefore, it is reasonable to think that some factor(s) determining the enzyme specificity must be present in the ␤-subunit.
The ␤-subunit proteins of MS2 and GA replicases are approximately 10% shorter than those of Q␤ and SP replicases, with most of the extra amino acids being located in the carboxylterminal region (4). Furthermore, the carboxyl termini of the ␤-subunit are more divergent in sequence (Fig. 1). It is therefore likely that this region may contribute to template discrimination.
To clarify in detail the physiological role of the carboxylterminal part of the ␤-subunit in the Q␤ replicase reaction, we constructed a series of plasmids harboring the ␤-subunit gene with deletions at its 3Ј-terminal end and examined the effects of the deletions on the enzyme functions. The carboxyl-terminal 18 amino acid residues were found to be dispensable for the replicase reaction, but a larger deletion made the enzyme defective.

EXPERIMENTAL PROCEDURES
Bacteria, Phage, and Plasmids-Escherichia coli A/ (sup pro/F ϩ ) and BE110 (sup ϩ rel-1 T2 r tonA22 phoA4 Hfr) were used as indicators. E. coli 594/FЈ lacI q (sup ϩ gal Str r recA44/FЈlacI q lac ϩ proAB) was used as a recipient for transformation and for growing phages. Phage Q␤sus51 was described previously (13). Plasmid pRQ1, which contains the wild-type ␤-subunit gene of Q␤ replicase, was used to construct plasmids carrying various sizes of the replicase gene lacking the 3Јterminal region. Plasmid pUC-MDV-LR carrying the segment corresponding to MDV-poly(ϩ) RNA was described previously (10). MDVpoly(ϩ) RNA is 244 nucleotides long and is a derivative of naturally occurring MDV-1(ϩ) RNA (14), for which Q␤ replicase exhibits strong template activity. Plasmid ZpQ␤-32 (15)  Construction of Plasmids Carrying the ␤-Subunit Gene-Plasmids carrying the ␤-subunit gene, of which the 3Ј-terminal end was shortened to various extents, were constructed using the polymerase chain reaction method. Oligonucleotide RQ1-f2 has a nucleotide sequence corresponding to positions 3328 -3351 of the Q␤ cDNA sequence including the cleavage site for the restriction enzyme, XhoI, and was used as a forward primer (Table I). The oligonucleotides listed in Table I were used as reverse primers, containing an artificial stop signal leading to premature termination of ␤-subunit protein synthesis and having the site for BglII for inserting the DNA fragment into the vector plasmid. Amplification of the DNA fragments was carried out according to the manufacturer's instructions. The amplified DNA fragments were digested with XhoI and BglII and then ligated to plasmid pVAR-AD6 DNA (9), which harbors the ␤-subunit gene lacking the 3Ј-terminal part and was previously cleaved with XhoI and BamHI. The nucleotide sequences of the amplified DNA regions were analyzed by the chain termination method (16) after the DNA fragments had been subcloned into the M13 phage. The plasmids constructed were designated according to the oligonucleotides used as the reverse primers.
In Vivo Complementation Analysis-E. coli 594/FЈ lacI q cells carrying the plasmids were grown at 37°C in YT broth (0.8% tryptone (Difco), 0.5% yeast extract, 0.5% sodium chloride) containing 10 mM CaCl 2 to a cell density of 3 ϫ 10 8 cells/ml, at which time phage Q␤sus51, which is defective in the ␤-subunit gene, was added at a multiplicity of infection of 0.1. After 10 min at 37°C, the cells were spun down and resuspended in an equal volume of YT broth. The cell suspension was then diluted 10 4 -fold with YT broth containing 2 mM isopropyl-1-thio-␤-D-galactopyranoside, and the incubation was continued at 37°C for 90 min, at which time chloroform was added. An aliquot (100 l), after appropriate dilution, was immediately mixed with a large excess of a stationary culture of E. coli A/ or BE110 and then plated onto a YT broth-agar plate.
Preparation of Cell Lysates-E. coli 594/FЈ lacI q cells carrying the plasmids were grown in 5 ml of YT broth containing 50 g/ml ampicillin to a cell density of 2 ϫ 10 8 cells/ml, at which time 2 mM (final concentration) of isopropyl-1-thio-␤-D-galactopyranoside was added to the medium, and the incubation was continued for another hour. Cell lysates were prepared according to Ball and Kaesberg (17). Aliquots (5 l) of the lysates were analyzed by SDS-polyacrylamide gel electrophoresis to demonstrate that a similar amount of ␤-subunit protein was present in each lysate (Fig. 2). For the gel retardation assay, the lysates were centrifuged at 30,000 ϫ g for 30 min at 4°C, and the supernatants were used.
Assay for Polymerase Activities Using Cell Lysates-The wild-type and mutant polymerase activities were measured using poly(rC), MDVpoly(ϩ) RNA, or Q␤ RNA as described previously (10), with slight modifications. The standard reaction mixtures (50 l) each contained 125 mM Tris-HCl (pH 8.0), 25 mM 2-mercaptoethanol, 10 mM MgCl 2 , 50 mM phosphoenol pyruvate, 10 g/ml pyruvate kinase, 74 units/ml DNase I, 10 g/ml rifampicin, and 5 l of cell lysate. For the poly(rG) experiment, the mixture also included 0.4 mM GTP, 20 g/ml poly(rC), and 740 kBq/ml [␣-32 P]GTP. The MDV-poly RNA or Q␤ RNA reaction mixture contained 0.8 mM each ATP, CTP, and GTP, 0.1 mM UTP, 20 g/ml MDV-poly(ϩ) RNA or 68 g/ml Q␤ RNA, and 1.48 MBq/ml [␣-32 P]UTP. Incubation was performed at 35°C for 15 min (MDV-poly RNA polymerase assay) or at 37°C for 60 min (Q␤ RNA polymerase assay). Aliquots (10 l) of the mixtures were applied to 3MM Whatman filter paper discs. Incorporated radioactivity was determined by liquid scintillation counting (10). In control reactions, 100 ng of purified Q␤ replicase was exogeneously added to the lysates of host cells carrying pUC8. From the control assay results, it was determined that 5 l of the lysates of cells carrying pRQ1 contained the equivalent of 271 ng (Q␤ RNA polymerase activity) or 247 ng (MDV-poly RNA polymerase activity) of pure Q␤ replicase.
In this paper, MDV-poly(ϩ) RNA represents an RNA that was transcribed from pUC-MDV-LR DNA by T7 RNA polymerase in vitro.
Northern Blot Analysis-E. coli 594/FЈ lacI q cells were grown and infected as described under "In Vivo Complementation Analysis" except that the phage was added at a multiplicity of infection of 3. After 10 min

5Ј-CGAGATCTATAGAGACGCAACCTTC-3Ј
Leu-585 a Underlined sequences indicate the termination signals in the reverse primers. b Numbers indicate the amino acid positions in the wild-type ␤-subunit protein, where the protein synthesis was terminated artificially at the mutant gene.
c An Ala or Leu residue was substituted for the Ile-564 residue of the DI564 protein.
at 37°C, the cells were spun down and resuspended in an equal volume of YT broth containing 2 mM isopropyl-1-thio-␤-D-galactopyranoside, and then the incubation was continued for the indicated times. RNAs were extracted with phenol and precipitated with ethanol in the presence of 0.8 M LiCl. The pellets were washed twice with 70% ethanol and then dissolved in distilled water. A 5-g sample of the extracted RNA was denatured with glyoxal and dimethyl sulfoxide and then applied to a 1.0% agarose gel. Electrophoresis, transfer to a nylon membrane (Biodyne A, Pall Ultrafine Corp.), and hybridization were performed as described previously (13).
A 1.2-kilobase pair XhoI-BamHI fragment of ZpQ␤-32 DNA, which lies within the maturation (A2) gene, was labeled using E. coli DNA polymerase I (Klenow fragment), random oligonucleotide primers (nonamers), and [␣-32 P]dCTP and then was used as the hybridization probe to detect Q␤ viral RNA.
Purification of the Replicase-The wild-type and DS567 replicases were purified from E. coli 594/FЈ lacI q cells carrying the plasmids according to Kajitani and Ishihama (19). For the Q␤ RNA polymerase assay, the standard reaction mixtures (50 l) each contained 80 mM Tris-HCl (pH 7.5), 12 mM MgCl 2 , 12 mM 2-mercaptoethanol, 0.8 mM each ATP, CTP, and GTP, 0.1 mM UTP, 74 kBq of [␣-32 P]UTP, and 200 ng of the replicase. Incubation was performed at 37°C for 15 min. Incorporated radioactivity was determined as described under "Assay for Polymerase Activities Using Cell Lysates."

RESULTS
Phage Growth in Cells Carrying the Plasmids-We first investigated whether or not the mutated ␤-subunit proteins could complement the activity of phage Q␤sus51 replicase, an amber mutant of the ␤-subunit gene. E. coli 594/FЈ lacI q cells carrying the mutant plasmids were infected with phage Q␤sus51, and then proliferation of the phage in the cells was examined.
As shown in Fig. 3A, as long as the amino acid deletion was less than 19 residues from the carboxyl-terminal end of the ␤-subunit protein, cells carrying each plasmid, pRQ1-DL585, -DQ578, -DV576, -DA572, and -DV570, produced as many progeny phages as did cells carrying the wild-type plasmid pRQ1 (Fig. 3A). However, when the deletion extended beyond the Val-570 residue, progeny production steadily decreased. Furthermore, when the deletion reached the Ile-564 residue, the phage production in cells carrying larger deletion plasmids was abolished completely.
Northern Blot Analysis-To investigate the cause of the poor phage production in cells carrying mutant plasmids (Fig. 3A), we extracted RNAs from phage Q␤sus51-infected cells, and subjected the RNAs to Northern blot analysis.
As shown in Fig. 4, the 32 P-labeled Q␤ cDNA probe strongly hybridized with Q␤ phage RNA (Fig. 4, lane a) but did not hybridize with cellular RNAs at all (Fig. 4, lane b), demonstrating that this cDNA probe specifically detected Q␤ viral RNA. In the control cells carrying pRQ1, viral RNA synthesis occurred at 20 min after the Q␤sus51 infection, and many viral RNAs accumulated during another 20 min of incubation (Fig. 4, lanes  i and j). On the other hand, in the cells carrying plasmid pRQ1-DA565 or -DS567, in which phage production was greatly reduced (Fig. 3A), viral RNA was found only a little even at 40 min after the infection (Fig. 4, lanes e-h). Since viral RNA synthesis did not occur in the cells carrying pUC8 (Fig. 4,  lanes c and d), the residual viral RNA syntheses found in the cells carrying the mutant plasmids were not due to the invad- ing phage genomes.
Also seen were the hybridization signals at the positions near to those of 16 and 23 S rRNAs (Fig. 4, lanes f and h-j). These signals may represent the premature transcripts of Q␤ RNA because such small RNAs were observed in Q␤ RNA-dependent RNA synthesis by wild-type or DS567 replicase (see Fig. 5C).
Q␤ RNA Polymerase Activity-Using lysates from cells carrying various mutant plasmids described above, we examined Q␤ RNA-dependent RNA polymerizing activity.
As shown in Fig. 3B, as long as the amino acid deletion was less than 18 residues, the polymerizing activity of lysates from the cells carrying the mutant plasmid was similar to that of the control cell lysate. However, when the deletion extended be-yond the Val-570 residue of the ␤-subunit, despite the fact that the cells expressing the mutant replicases supported poor growth of the invading Q␤sus51 phage (Fig. 3A), lysates of the cells carrying pRQ1-DR569, -DS568, -DS567, or -DC566 synthesized much more RNA than did lysate from cells harboring pRQ1. These results suggest that those deletions did not affect the polymerizing activity of the replicase. When 25 or more amino acids were deleted from the carboxyl terminus of the ␤-subunit protein, polymerizing activity was abolished. In addition, lysates of cells carrying pUC8 exhibited little RNA polymerizing activity.
To further analyze the characterization of the mutant replicase, we purified DS567 replicase and compared the Q␤ RNAdependent RNA synthesis by wild-type and DS567 replicases. As shown in Fig. 5A, when 16 ng of the template Q␤ RNA was added to the reaction mixture, DS567 replicase synthesized RNA in a similar manner and amount to wild-type replicase. However, when 1 g of the template RNA was incubated with DS567 replicase, incorporation of [ 32 P]UTPs was about 2-fold of that using wild-type replicase. Fig. 5B shows the relationships between Q␤ RNA-dependent synthesis and the molar ratio of Q␤ RNA to replicase (Q␤ RNA/replicase). Since Q␤ replicase binds to Q␤ RNA at three sites (5), an excess amount of Q␤ RNA causes binding of Q␤ replicase to two or three Q␤ RNA molecules, resulting in inhibition of Q␤ RNA synthesis (20). Under our experimental conditions, Q␤ RNA-dependent synthesis was found to be maximum at a molar ratio of about 0.3 or 0.5 for the wild-type or DS567 replicase reaction, respectively. When the ratio was higher than 0.5, DS567 replicase synthesized RNA twice as much as wild-type replicase did, which was consistent with the results in Fig. 4A. When the 32 P-labeled RNAs synthesized by wild-type or DS567 replicase at a molar ratio of 0.5 were compared by an agarose gel elec-  (lanes b and d). Equal amounts of radioactive RNA (1,100 cpm) synthesized by DS567 (lanes a and b) or wildtype (lanes c and d) replicase at the molar ratio of 5 were denatured and applied to each lane of a 1% agarose gel. After electrophoresis, RNAs were transferred to a membrane (Biodyne A). The positions of Q␤ RNA, 16 and 23 S ribosomal RNAs were determined as in Fig. 4. trophoresis, most of the 32 P-labeled RNAs were found to be similar in size to Q␤ viral RNA (Fig. 5C). These results indicate that the polymerization activity of DS567 replicase did not get damaged and suggest that DS567 replicase had an extra ability to bind to Q␤ RNA.
Template Specificity of the Mutant Replicases-It was puzzling that lysates of cells carrying pRQ1-DC566, -DS567, -DS568, and -DR569 exhibited higher polymerizing activity in vitro with the Q␤ RNA template but that these cells were unable to support effectively the proliferation of the invading phage, Q␤sus51.
Taking account of the results in Fig. 5, the mutant replicases are assumed to have more relaxed template specificity than the wild-type replicase, so thereby they neglect Q␤ RNA synthesis in cells infected with phage Q␤sus51. Therefore, we examined the polymerizing activities of cell lysates using cellular RNAs as templates. As shown in Table II, lysates of cells carrying pRQ1-DC566, -DS567, -DS568, or -DR569 incorporated more [ 32 P]UTPs than the control cell lysates carrying pRQ1 when ribosomal RNA or total cellular RNA was used as template. In contrast, lysates of cells carrying pRQ1-DA572 or -DV576, which exhibited lower Q␤ RNA polymerase activity in vitro but supported progeny production in Q␤sus51-infected cells, had lower polymerization activity than the control lysate in the cellular RNA-dependent system.
To further determine the cause of the difference between the in vivo and in vitro replicase reactions, we compared the template specificity of purified wild-type and DS567 replicases. Q␤ replicase transcribes closely related phage RNAs but not other RNAs such as ribosomal RNAs and unrelated phage RNAs (5). As shown in Table III, when SP RNA, a closely related phage RNA to Q␤ RNA, was also used as a template, DS567 replicase incorporated more [ 32 P]UTPs than the wild-type replicase, as in the case of Q␤ RNA template. DS567 replicase also exhibited much higher RNA polymerizing activity for 16 and 23 S ribosomal RNAs or total cellular RNA as template, confirming the results of Table II. These results indicate that DS567 replicase relaxed specificity for not only Q␤ RNA template but also cellular RNA templates.
MDV-poly RNA and Poly(rG) Polymerase Activities-Since Q␤ replicase lacking S1 and HF-I does not transcribe Q␤ RNA but can use MDV-1 RNA and poly(rC) as template, we examined the effects of deletions on MDV-poly RNA and poly(rG) polymerase activities. The MDV-poly(ϩ) RNA-dependent polymerizing activity of cell lysates changed in a similar way to Q␤ RNA-dependent polymerizing activity as the amino acid deletion extended into the inner part of the ␤-subunit protein (Fig. 3B). When 25 residues were deleted, lysates of cells expressing mutant replicases with larger deletions no longer showed the polymerizing activity.
The poly(rC)-dependent polymerizing activity of cell lysates carrying pRQ1-DI564 decreased by 75% compared with the control cell lysates, as in the case of the MDV-poly(ϩ) RNA template (Table IV). Furthermore, replacement of the carboxylterminal Ile residue of DI564 replicase with Ala or a chemically similar Leu residue reduced the poly(rG) and MDV-poly(ϩ) RNA polymerizing activities to background levels. These results indicate that the sequence from Ile-564 to Ala-565 of the ␤-subunit protein is important for both the MDV-poly RNA and poly(rG) polymerase activities.
MDV-poly(ϩ) RNA Binding Activity-Q␤ replicase binds to the central region of MDV-1(ϩ) RNA, in which nucleotides 81-126 are identical with nucleotides 84 -129 of Q␤ minusstrand RNA (21). To determine whether or not mutant replicases without MDV-poly RNA polymerizing activity can bind to a template, we examined a gel retardation assay using 32 P-endlabeled MDV-poly(ϩ) RNA. When the lysates of cells carrying pRQ1-DA565, pRQ1-DC566, or pRQ1 were incubated with [ 32 P]MDV-poly(ϩ) RNA and then electrophoresed on a polyacrylamide gel, the mobility of [ 32 P]MDV-poly(ϩ) RNA was decreased (Fig. 6, lanes c and g-i). In contrast, in the case of lysates of cells carrying pRQ1-DI564, which exhibited poor  polymerizing activity (Fig. 3B and Table IV), only a small amount of [ 32 P]MDV-poly(ϩ) RNA migrated slowly (Fig. 6, lane f), indicating that the binding of MDV-poly(ϩ) RNA to DI564 replicase was greatly impaired. Furthermore, in the case of pRQ1-DQ562 or -DY563, a mobility shift was not observed anymore ( Fig. 6, lanes d and e). DISCUSSION Q␤ replicase distinguishes its own Q␤ RNA and closely related viral RNAs among natural RNAs. This discrimination involves the recognition of a specific secondary or tertiary structure of the template RNA. Q␤ RNA provides the replicase with three binding sites (S-site, the initiation region of the coat gene; M-site, the internal region of the ␤-subunit gene; and the 3Ј-terminal region of Q␤ RNA) (3). However, little is known about the counterparts of the replicase molecule responsible for the enzyme specificity, such as template recognition or subunit assembly. In this paper, we propose that the carboxyl-terminal part of the ␤-subunit of Q␤ replicase is one of the candidates for functional domains determining the template recognition from the following results.
Q␤ RNA Polymerizing Activity-DS567 replicase with deletions of 21 amino acid residues at the carboxyl terminus had a relaxed template specificity for a related phage RNA and cellular RNAs as well as Q␤ RNA (Tables II and III, Fig. 5). Furthermore, lysates of cells expressing DR569, DS568, DS567, or DC566 replicase, which showed in vitro higher polymerizing activity with the Q␤ RNA template but failed to support progeny production in the cell, exhibited higher po-lymerizing activity in cellular RNA-dependent synthesis (Table  II). In contrast, lysates of cells expressing DA572 or DV576 replicase showed lower polymerizing activity than the control lysate in Q␤ RNA or cellular RNA-dependent synthesis (Table  II); however, these mutant proteins complemented the activity of phage Q␤sus51 replicase (Fig. 3A). These results suggest that in cells infected with phage Q␤sus51, a mutant replicase having a more relaxed template specificity than that of wildtype replicase engaged in transcribing cellular RNAs as well as Q␤ RNA and accordingly failed to perform the task of polymerizing Q␤ RNA, indicating that the accessibility of Q␤ replicase to cellular RNAs may be critical for phage growth.
MDV-poly(ϩ) RNA Polymerase Activity-DI564 replicase lacking the carboxyl-terminal 24 amino acid residues of the ␤-subunit protein showed only a little activity in the MDV-poly RNA polymerase assay (Fig. 3B). On removal of the terminal Ile residue from DI564 replicase, the resulting mutant, DY563, replicase lost the polymerizing activity (Table IV). Furthermore, DI564 replicase showed reduced binding ability to MDVpoly(ϩ) RNA, and DY563 replicase abolished it completely (Fig.  6). These results indicate that the failure of MDV-poly(ϩ) RNAdependent RNA synthesis by DI564 replicase was due to poor binding of the replicase to the RNA. Therefore, the sequence from Ile-564 to Ala-565 of the ␤-subunit protein was essential for Q␤ replicase to express MDV-poly RNA polymerase activity, particularly for its binding to MDV-poly(ϩ) RNA. The finding that the insertion of a pentapeptide at the Ala-565 residue abolished the MDV-1 RNA polymerizing activity of Q␤ replicase (6) agrees with our results.
Recently, Brown and Gold (22) have proposed a three-site model of Q␤ replicase, according to which two RNA binding sites, site I on the S1 subunit and site II on EF-Tu, are responsible for RNA binding, and the polymerase-active site on the ␤-subunit is for RNA synthesis. Site I binds class I RNAs that possess single-stranded regions containing a high fraction of A and C nucleotides such as Q␤ plus-strand RNA, and site II binds class II RNAs that have polypyrimidine tracts such as Q␤ minus-strand RNA or MDV-1 RNA. According to their model, destruction of site II or removal of EF-Tu from Q␤ replicase will result in the replicase without the MDV-1 RNA binding activity. Therefore, the data in Fig. 6 that show the importance of the Ile-564 and Ala-565 residues of the ␤-subunit protein in MDV-poly(ϩ) RNA binding suggest that site II on EF-Tu interacts with these amino acid residues to form an active Q␤ replicase molecule.
Our present results indicate that a change in the carboxylterminal structure of the ␤-subunit protein of Q␤ replicase could cause relaxation of the template specificity and that this part of the ␤-subunit is responsible for recognizing MDVpoly(ϩ) RNA. The carboxyl-terminal region of the ␤-subunit protein is heterologous in RNA coliphages. Phage RNA replicase is thus assumed to have evolved its template specificity in part by altering the carboxyl-terminal structure of the ␤-subunit and simultaneously by keeping the accessibility of the replicase to cellular RNAs at a low level.