ABSTRACT
DNA polymerases (DNAPs) recognize 3′ recessed termini on duplex DNA and carry out nucleotide catalysis. Unlike promoter-specific RNA polymerases (RNAPs), no sequence specificity is required for binding or initiation of catalysis. Despite this, previous results indicate that viral reverse transcriptases bind much more tightly to DNA primers that mimic the polypurine tract. In the current report, primer sequences that bind with high affinity to Taq and Klenow polymerases were identified using a modified systematic evolution of ligands by exponential enrichment (SELEX) approach. Two Taq-specific primers that bound ∼10 (Taq1) and over 100 (Taq2) times more stably than controls to Taq were identified. TaqI contained 8 nucleotides (5′-CACTAAAG-3′) that matched the phage T3 RNAP “core” promoter. Both primers dramatically outcompeted primers with similar binding thermodynamics in PCRs. Similarly, exonuclease− Klenow polymerase also selected a high-affinity primer that contained a related core promoter sequence from phage T7 RNAP (5′-ACTATAG-3′). For both Taq and Klenow, even small modifications to the sequence resulted in large losses in binding affinity, suggesting that binding was highly sequence specific. The results are discussed in the context of possible effects on multiprimer (multiplex) PCR assays, molecular information theory, and the evolution of RNAPs and DNAPs.
IMPORTANCE This work further demonstrates that primer-dependent DNA polymerases can have strong sequence biases leading to dramatically tighter binding to specific sequences. These may be related to biological function or be a consequence of the structural architecture of the enzyme. New sequence specificity for Taq and Klenow polymerases were uncovered, and among them were sequences that contained the core promoter elements from T3 and T7 phage RNA polymerase promoters. This suggests the intriguing possibility that phage RNA polymerases exploited intrinsic binding affinities of ancestral DNA polymerases to develop their promoters. Conversely, DNA polymerases could have evolved from related RNA polymerases and retained the intrinsic binding preference despite there being no clear function for such a preference in DNA biology.
INTRODUCTION
Most DNA polymerases (DNAPs) recognize 3′ recessed termini on double-stranded nucleic acid and use this feature as the priming point for nucleotide catalysis. Unlike promoter-dependent RNA polymerases (RNAPs), it is thought that sequence-specific information from the duplex region plays only a small role, if any, in polymerase binding and catalysis. Despite this, retroviral reverse transcriptases (RTs) bind much more tightly to purine-rich DNA-DNA duplexes with primers resembling their polypurine tract (PPT) RNA sequences (5′-AAAAGAAAAGGGGGG-3′ for HIV-1) (1, 2). The PPT is resistant to RNase H degradation, which allows its use as a primer for second-strand DNA synthesis (3–6). In addition, it is also a more efficient primer than random RNA sequences, suggesting a unique interaction with RT (7–9). The finding that PPT-like sequences also induce high-affinity binding to RT further elucidates the PPT's preferential usage for second-strand priming and demonstrates that DNAPs can have strong sequence binding preferences.
The sequence preferences for RTs were uncovered using a modified systematic evolution of ligands by exponential enrichment (SELEX) protocol that allows the selection of tight-binding primer-template sequences (referred to as primer-template SELEX [PT-SELEX] in this paper) (1, 2). PT-SELEX starts with a pool of duplex nucleic acids that have a 4-base 5′ overhang creating a 3′ recessed end within a region of random sequence nucleotides. Polymerases bind preferentially to those sequences that induce stronger binding, which get selected in the subsequent rounds of PT-SELEX.
Given that viral RTs have a biologically relevant sequence binding preference, it is possible that other primer-template-utilizing polymerases also have preferences for specific sequences. Though diverse, DNAPs can be grouped by sequence and structural homology into seven families (A, B, C, D, X, Y, and RT), all of which share several analogous regions that are necessary for their function, including the catalytic palm domain, the fingers domain, the thumb domain (which helps position the nucleic acid template), and a two-metal-ion binding site in the catalytic cleft (10, 11).
The 832-amino-acid DNA polymerase from the bacterium Thermus aquaticus, known as Taq polymerase (Taq pol), is perhaps the most commercially important polymerase. Taq polymerase, which is classified in family A, is thermostable at temperatures that would denature many other proteins: its optimal temperature range is 75 to 80°C, and it can withstand temperatures near boiling (at 97.5°C, the enzyme has a half-life [t1/2] of 9 min) (12). This has made it ideal for PCR-based applications, which repeatedly cycle through ≥90°C temperatures (13).
Given the commercial importance of Taq polymerase, it represents an interesting target for investigation of sequence preferences. In this report, we show that Taq has strong sequence binding preferences and shares a preference for sequences that resemble phage RNAP promoters. Primers for PCR mixtures containing Taq high-affinity binding sequences had a dramatic competitive advantage over other primers, resulting in a strong bias for the production of PCR products specified by these primers. Further, we found that another commercially important family A polymerase, the Klenow fragment of Escherichia coli DNAP I, also has a binding preference for specific sequences, and these sequences also resemble phage RNA polymerase promoter sequences. These results suggest possible evolutionary and structural relationships between promoter- and primer-dependent polymerases that will be discussed in this paper. Further, the results demonstrate the strong competitive advantage of selected sequences in PCRs and indicate that primer bias not only can occur due to thermodynamic nucleic acid binding advantages of some primers (14–16) but also may result from specific sequences binding Taq with higher affinity. These findings have possible ramifications for quantitative PCRs and PCR-based protocols.
RESULTS
Taq polymerase selected a high-affinity sequence containing a conserved region of the T3 RNA polymerase promoter and a second unrelated sequence.PT-SELEX experiments with Taq were carried out for 7 rounds using the approach shown in Fig. 1 (see also references 1 and 2). Material from rounds 6 and 7 bound Taq by gel shift analysis nearly equivalently, indicating that selection was essentially complete. Twelve sequences were isolated from the round 7 pool, of which two different sequence groups were identified (Fig. 2). In group 1, five sequences were identical except for a single nucleotide substitution in one isolate (Fig. 2B). Interestingly, 4 sequences (Fig. 2B, sequences 1, 2, 4, and 5) contained a region corresponding to bases −7 to +1 of the phage T3 RNAP promoter region (5′-CACTAAAG-3′, underlined in the consensus sequence at the bottom of Fig. 2B), while the fifth had a one-nucleotide change in that region (Fig. 2B, sequence 3, 5′-CCCTAAAG-3′). This region of the T3 promoter, which constitutes a core sequence that is highly conserved between T3, T7, and SP6 RNAPs (T3Pcore) (17), was contained fully within the duplex DNA (4 bp upstream from the 3′ primer terminus, which is denoted by the darker-gray shade in Fig. 2). Based on these 5 sequences, a representative consensus sequence (identical to 4 of the 5 sequences in the set) was produced for further testing and is referred to as Taq1 (Table 1). The second group of three sequences identified from the 12 recovered sequences shared an identical 8-base sequence (5′-AACGTGCC-3′) and were closely related in other regions (Fig. 2C). Two of the sequences were identical except for a single nucleotide change. Unlike the first set of sequences, the 8-base sequence in this second set had no similarity to any known biologically relevant motifs. A sequence that was identical to the recovered sequence 3 in Fig. 2C was produced for further testing and is referred to as Taq2 (Table 1). An alignment of Taq1 and other selected sequences with the T3 and T7 RNAPs is shown in Fig. 3. Note the strong homology of the T3 RNAP with Taq1 in Fig. 3A. In addition to the identical 8-base stretch noted above, homology to a downstream “AGA” sequence of the T3 and T7 promoters was also apparent. Taq2 showed no significant homology to either RNAP.
Primer-template SELEX (PT-SELEX) protocol for selecting primer-template sequences that bind Taq and Exo−Kl with high affinity. The preparation of the top construct is described in Materials and Methods. Refer to Materials and Methods for a detailed description of the selection process for Taq and Exo−Kl polymerases.
Alignment of recovered round 7 Taq-selected sequences. Nucleotides that match the consensus, located beneath each alignment, are represented by a dot. Sequences are from the random region of the primer strand, and the four bases corresponding to the single-stranded template overhang are shaded darker gray (Fig. 1). Note that those sequences are the complements of the bases that were present in the template overhang (Table 1). Of the 12 sequences isolated (A), two general motifs were observed. Group 1 (B) contained a region that matched the initiation domain (underlined) of the T3 RNAP promoter. Group 2 (C) did not contain a recognizable biologically relevant sequence. Sequences that were tested further are represented by sequence 1 in panel B (Taq1) and sequence 3 in panel C (Taq2). Sequences were aligned using MacVector. Codes: M is A or C; R is A or G; W is A or T; Y is C or T; S is C or G; K is G or T.
Selected SELEX sequences from Taq and Exo−Kl
Clustal alignment of the selected sequences from the Taq SELEX (Taq1 and Taq2) and the exonuclease− Klenow SELEX (Exo−Kl-1) with T3 (A) and T7 (B) RNAP promoter sequences. T7 and T3 RNAP sequences are extended 23-nucleotide versions of the promoter regions illustrated in Fig. 7. The binding and initiation regions from that figure are highlighted above the illustrations in yellow and gray, respectively. The primer strand of the 25-nucleotide regions of Taq1, Taq2, and Exo−Kl-1, derived from the random nucleotides in the starting material, were used in the alignments. Dashes are gaps in the alignments, while dots indicate the same nucleotide as that in the reference sequence at that position.
Determination of important sequences for Taq binding in the selected material.Both Taq1 and Taq2 bound much more tightly than the starting material (control) in gel shift assays (Fig. 4). Taq2 appeared to bind modestly better than Taq1 in this environment. As these experiments used relatively large amounts of nucleic acid starting material (2 nM) and the conditions were not optimal for Taq binding, the gels were used only to compare binding to different constructs (filter binding assays for some constructs were also performed [see below]). Apparent equilibrium dissociation constants (referred to as Kd,app,gelshift; see Materials and Methods) were determined for comparisons and are shown in Table 1. Taq1 and Taq2 had Kd,app,gelshift values of 36 ± 6 and 6.3 ± 0.1 nM, respectively, while the control (starting material in the PT-SELEX) bound too weakly to estimate a value.
Gel shift assay of Taq1 and Taq2 constructs, as well as random starting material by Taq pol. The constructs are shown in Table 1. Taq1 and Taq2 were selected from round 7 of SELEX (Fig. 1). In each gel, the concentrations of Taq pol were 0, 6.25, 12.5, 25, 50, 100, and 200 nM. The apparent affinities for the constructs using gel analysis (referred to as Kd,app,gelshift) were 36 ± 6 nM and 6.3 ± 0.1 nM for Taq1 and Taq2, respectively (averages from 3 experiments ± SD). The starting material bound too weakly to determine a Kd value.
Mutational analysis was performed on the Taq1 and Taq2 sequences to determine what regions of the sequences were important for high-affinity binding to Taq (Tables 2 and 3 show data obtained for Taq1 and Taq2, respectively). In general, the results showed that both the absolute sequence and the location of the conserved regions of Taq1 and Taq2 were necessary for tight binding. To test the role of the T3Pcore sequence for binding of Taq to Taq1, a construct was created with substitutions at primer nucleotides flanking the T3Pcore (“Modified sequences surrounding the T3Pcore” in Table 2), as well as a second construct with a completely modified T3Pcore sequence (“Modified T3Pcore” in Table 2). Using gel shift analysis, both constructs bound less tightly than Taq1 to Taq, but the modified T3Pcore construct bound much less tightly and was essentially equivalent to the control, while the construct with an intact T3Pcore but modified surrounding sequences bound modestly less strongly. This indicated that the T3Pcore was an important factor in the enzyme's affinity for the construct but that other sequences surrounding this region enhanced binding to a smaller extent. This was not surprising, as the conserved region of Taq1 extended beyond the eight-nucleotide T3Pcore bases (Fig. 2). By mutating 4-bp segments of the T3Pcore, it was found that the bases matching from −3 to +1 (5′-AAAG-3′, in the primer strand) in the T3Pcore were the largest contributors to tight binding (“4-bp modifications B” in Table 2); mutation of these bases disrupted binding, while mutating the −7 to −4 bases (5′-CACT-3′, in the primer strand, “4 bp modification A” in Table 2) did not significantly affect binding. Intriguingly, sometimes small changes to the T3Pcore were more deleterious to binding than large ones: a single change of the −2 T3Pcore base (to mimic the phage T7 promoter core sequence, “T3Pcore to T7Pcore,” in Table 2) completely disrupted strong binding. Also, shifting the position of the core sequence even by a single base relative to the 3′ primer terminus completely abolished tight binding in the gel shift assay (“−1 shift” in Table 2). Interestingly, Taq polymerase bound a construct that contained the entire T3 RNAP promoter (Table 2, “full T3 RNAP promoter sequence”) as poorly as it did the random control, suggesting that a construct with the entire promoter sequence is in a conformation that disrupts high-affinity Taq binding, despite the presence of the T3Pcore.
Effects of modifying Taq1 sequence on affinity for Taq polymerase
Effects of modifying Taq2 sequence on affinity for Taq polymerase
Unlike Taq1, all tested modifications of Taq2 significantly disrupted binding (Table 3). Modification of the 8-bp region found in all three sequences (underlined in the Taq2 sequence in Table 3) reduced binding ∼8-fold, as did modification of the bases upstream and downstream of this region. Although some binding enhancement relative to the control was observed when only upstream bases were modified (“Modified noncore/nonoverhang nucleotides” in Table 3). As with Taq1, even a one-base shift of the sequence relative to the 3′ primer terminus completely abolished tight binding (“−1 shift” in Table 3).
Filter binding assays show that Taq1 and Taq2 bind extremely tightly to Taq in comparison to other sequences.Since protein-nucleic acid complexes for some proteins may be weakened in gel environments, other parameters of these sequences were evaluated using nitrocellulose filter binding assays (Table 4). Consistent with this, even the random starting material bound to Taq polymerase strongly in filter binding assays with a measured Kd of 14 ± 8 pM at 23°C. The Kd of Taq1 and Taq2 was less than a few picomolar under these conditions, making it too small to accurately calculate in this assay. Binding half-lives were also calculated; at 23°C, Taq1 and 2 bound Taq with a half-life of greater than 8 h, and in fact, almost no dissociation of the sequences from Taq was detected after 8 h for either sequence. In contrast, the random starting material had a binding half-life of ∼20 min. At 60°C, all the sequences showed more-rapid dissociation from Taq, consistent with previous results showing that binding of Taq to primer-templates is less stable at higher temperature (18, 19). The random control in this case bound with a half-life of ∼4.4 ± 1.2 min, while Taq1 showed about 10-fold-more-stable binding. Taq2 bound about 100 times more stably than the control at this temperature, consistent with tighter binding than Taq1 in the gel shift assay (Table 1 and Fig. 3). Since Taq2 was G-C rich near the 3′ end of the primer strand and therefore more stable at higher temperature, it was possible that this, rather than a sequence-specific intrinsic binding advantage, was responsible for the extremely stable binding at 60°C. However, the Taq2 “−1 shift” sequence (see above and Table 3), which differs from Taq2 by just a single nucleotide, bound poorly in gel shift experiments (Table 3) and had a half-life at 60°C of only 3.5 ± 1.5 min, essentially identical to the random sequence control half-life (Table 4). This suggests that the relatively G-C-rich nature of Taq2 plays no role in the observed results. Technical issues made it impossible to accurately measure Kd values by filter binding at 60°C, but the half-life measurements indicate that the selected sequences bind much more tightly than random primer-template sequences, even at an elevated temperature.
Binding parameters at 23°C and 60°C for selected Taq sequences
The selected Taq1 and Taq2 sequences dramatically outcompete other sequences in PCRs.The isolation of high-affinity binding sequences provided an opportunity to test the potential effects of these sequences on PCRs. Results indicated that 20-nucleotide primers mimicking Taq1 (5′-CCAGTCCACACTAAAGCATA-3′) and Taq2 (5′-CCCAATTTGCGAACGTGCCT-3′) dramatically outcompeted other primers with similar thermodynamic binding properties in PCRs (see Fig. S1 and S2 and the accompanying descriptions in the supplemental material). The Taq1 and Taq2 primers were also effective in PCRs over a wide range of binding temperatures (see Fig. S3 in the supplemental material).
Exo−Kl polymerase selected a high-affinity sequence containing a conserved region from the T7 polymerase core sequence.A second commercially important bacterial DNA polymerase was also analyzed for sequence-specific binding preferences. In this case, the exonuclease-minus version of Klenow polymerase was used in order to avoid degradation of the starting material that would occur in the presence of Mg2+ with the exonuclease-containing enzyme. After 7 rounds of PT-SELEX (Fig. 1), the enriched pool bound Exo−Kl several times better than a random pool based on gel shift analysis (Kd,app,gelshift of 90 ± 10 nM versus >500 nM), suggesting that it contained high-affinity sequences. Round 6 material bound with approximately the same affinity as round 7 material, indicating that the selection was essentially complete. The round 7 pool was cloned and sequenced, and 12 isolates were recovered, of which 11 showed significant identity over a large region (Fig. 5). Interestingly, these 11 sequences contained a second BbsI site in an orientation opposite to the engineered site in the primer region (Fig. 1 and 5A). We later found that due to incomplete cleavage by the restriction enzyme, this allowed a nick on each strand that resulted in a 9-nucleotide overhang on the 5′ end of the template instead of the intended 4-nucleotide overhang (generating 36 nucleotide primer/45 nucleotide template pairs) (Tables 1 and 5). Subsequent testing revealed that reducing the length of the 5′ overhang by more than 3 nucleotides significantly disrupted binding (“36-nt primer-42-nt template” in Table 5), suggesting a strong selective pressure for the location of this second BbsI site. In fact, Exo−Kl was unable to preferentially gel shift any sequences that were tested containing 4-nucleotide overhangs of the type that the protocol was designed to produce (Fig. 1).
Alignment of recovered round 7 Exo−Kl selected sequences. (A) The starting material hybrid and original BbsI cut site (which generates a 41:45 nucleotide primer/template ratio) and the second selected cut site (which yields a 36:45 nucleotide primer/template ratio [as indicated by an asterisk] when combined with the first site) are shown. See Fig. 1 and Results for details. (B) The 11 sequences isolated from round 7 containing a portion of the T7 promoter region aligned using MacVector. Sequences are from the random region of the primer strand, and the nine bases corresponding to the single-stranded template overhang are shaded in darker gray. Note that those sequences are the complement of the bases that were present in the template overhang (Table 1). A consensus sequence is shown at the bottom with the T7 promoter element underlined. The first sequence corresponds to the Exo−Kl-1 sequence that was examined further.
Effects of modifying Exo−Kl-1 sequence on affinity for Exo−Kl polymerase
Intriguingly, 7 of the cloned round 7 sequences contained a region identical to a part of the T7 phage RNAP consensus promoter sequence (bases −6 to + 1 of the T7 polymerase promoter core sequence [T7Pcore] [17], 5′-ACTATAG-3′), while three others contained only a single nucleotide difference in that region (Fig. 5B). Note that the Taq1 sequence above contained a related region from the T3 promoter (nucleotides −7 to +1). In the phage RNAP promoters, this region is highly conserved between different RNAPs (a single base differentiates the T3 promoter from T7 and SP6); bases −4 to +3 are where a single-stranded bubble forms for the initiation of transcription by phage RNAP (17). The T7-like sequences selected by Exo−Kl contained four of the bases (−6 to −2 of the T7Pcore) in the duplex region and three bases (−2 to +1 of the T7Pcore) in the single-stranded template overhang of the primer-template hybrid. A consensus sequence, constructed using the most common bases from the 11 highly similar selected sequences (Fig. 5B, bottom reference sequence), was used for further testing to determine which bases contribute most to selective binding. An alignment of that sequence (named Exo−Kl-1) with the T7 and T3 RNAPs is shown in Fig. 3. Other than the 7 nucleotides from −6 to +1 noted above, there was no significant homology with other regions of the T7 or T3 RNAP.
In gel shift assays, Exo−Kl-1 bound with a Kd,app,gelshift of ∼125 nM to Exo−Kl polymerase, comparable to the overall Kd,app,gelshift of the round 7 material (Fig. 6 and Table 1). Notably, in gel shifts conducted in the absence of Mg2+, this sequence also bound Klenow wild-type polymerase, which has no mutation to eliminate the 3′→5′ exonuclease activity. Binding to Exo−Kl was essentially identical for Klenow and Exo−Kl, indicating that the mutation in Exo−Kl played no role in the sequences that were selected in the SELEX protocol (data not shown).
Gel shift of Exo−Kl-1, as well as random starting material by Exo-minus Klenow polymerase. The Exo−Kl-1 construct was selected from round seven of PT-SELEX and is shown in Table 1. The concentrations of Exo−Kl polymerase were 0, 25, 50, 100, 200, 400, and 800 nM for the starting material and 0, 3.1, 6.3, 12.5, 25, 50, 100, 200, 400, and 800 nM for Exo−Kl-1. The apparent affinity for the Exo−Kl-1 construct using gel analysis (referred to as Kd,app,gelshift) was 125 ± 25 nM (average from 3 experiments ± SD). The starting material bound too weakly to determine an affinity.
Determination of important sequences for Exo−Kl binding in the selected material.Mutational analysis was performed on the Exo−Kl-1 sequence to determine which bases were most responsible for the specific binding of Exo−Kl (Table 5; see also Fig. S4 in the supplemental material for an example of a gel shift experiment). In addition to constructs with template truncations (described above), which revealed the enzyme's preference for 9-nucleotide overhangs, two other major types of mutations were made: base substitutions and shifting of the primer-template sequences.
Modifying the 7 bases that made up the T7Pcore disrupted tight binding to Exo−Kl (Table 5, modified T7Pcore); as this sequence was found in most of the recovered sequences, the result is not surprising and implies that these bases contribute significantly to the specific interaction with the enzyme. Modifying the 9-nucleotide single-stranded template overhang sequence (which contains 3 nucleotides of the T7Pcore) (Table 5, modified template overhang) modestly decreased enzyme binding, while modifying the bases 5′ of the T7Pcore had a similar effect (Table 5, 5′ modified non-T7 sequence). Together, these results suggest that the four bases of the T7Pcore in the duplex region are the major contributors to tight binding but are aided by sequences upstream and downstream. This agrees with the observation that the region of conservation among recovered sequences is larger than just the −6 to +1 T7Pcore (Fig. 5B).
To determine if the context of the sequence relative to the primer terminus was important, constructs in which the entire random-region-derived sequences were shifted further in (−1, −4, −8) or out (+1) of the duplex region were created. While all shifts decreased binding, the −1 shift (which contained 5 bp of the T7Pcore in the duplex region and 2 nucleotides in the single-stranded region) was tolerated best; other shifts were more deleterious to binding, demonstrating that, like the selected Taq sequences, sequence context was also integral to tight binding.
DISCUSSION
In previous reports, we demonstrated that reverse transcriptases exhibit primer-template sequence-specific binding preferences for substrates that resemble the PPTs of the viruses (1, 2). The current report demonstrates that both Taq and Klenow polymerases also demonstrate primer-template sequence-specific binding preferences. In this case, sequences related to phage RNA polymerase promoters were selected with both enzymes (Fig. 7). Unlike the PPT-like sequences selected by HIV RT, there was no clear biological role for the specific sequences that were selected; however, they did reveal a novel relationship between phage RNAPs and specific bacterial DNA polymerases that could have functional, structural, and evolutionary implications.
T7 and T3 phage promoter sequences denoting regions of homology with Taq and Exo−Kl PT-SELEX selected sequences. The +1 bolded G residue indicates the start of transcription. Boxed regions denote sequences shared by the promoters and the selected tight-binding primer-template sequences for Exo−Kl (T7 promoter) and Taq (Taq1, T3 promoter) polymerases.
The evolutionary origin of single-subunit DNA-dependent RNAPs (ssRNAPs) is uncertain. In addition to phage, members of this class are found in chloroplasts, nucleus, and mitochondria. Given the presumptive bacterial origin of chloroplasts and mitochondria, it is conceivable that these enzymes have a bacterial/phage origin (this and other possibilities are discussed in detail in reference 20). However, primary sequence conservation to multisubunit RNA polymerases and DNA polymerases is low, which has made it difficult to trace the origin of this group. Despite low sequence homology, ssRNAPs share a structural core with other right-hand-shaped polymerases. Using the properties of the conserved amino acids in this core, Monttinen et al. (21) constructed a structure-based distance tree demonstrating that ssRNAPs are closely related to family A DNA polymerases, which include DNAP I and Taq. The striking structural similarities between T7 phage RNAP and the Klenow fragment of DNAP I and related polymerases (22–24), coupled with mutational analysis showing that deoxynucleoside triphosphate (dNTP) and ribonucleoside triphosphate (rNTP) usage can be altered within this group by a small number of amino acid changes (25), led Cermakian et al. (20) to hypothesize that ssRNAPs and DNAP I-like enzymes evolved by divergent evolution from a common ancestor. Our results further bolster these analyses by demonstrating the existence of a common sequence motif that can drive high-affinity binding. It is conceivable that during evolution, phage RNAPs exploited an intrinsic binding affinity of an ancestral polymerase to develop their promoters, while DNA polymerases retained the intrinsic binding preference despite there being no clear function for such a preference in DNA biology.
It is notable that the core region of phage RNAP promoters was selected by Taq and Exo−Kl (Fig. 7), as this is the most highly conserved region among phage RNAPs (17). Mutational analysis indicates that the core region (nucleotides −7 to +1) is pivotal for both initiation (maps most strongly to nucleotides −5 to +3) and RNAP binding (maps most strongly to nucleotides −12 to −5), while binding specificity is more strongly dictated by the less conserved upstream promoter sequences (17, 26, 27). The A-T-rich nature of the core region may also be important for allowing melting, which is observed in this region in crystal structures between T7 RNAP and promoter duplexes (22). Further, A-T-rich nucleic acid duplexes are known to be more flexible (28–31). As polymerases are well known to “bend” the primer-template duplex upon binding, this seems to be a possible role of the Taq1 sequence, given its positioning in the primer-template duplex (Fig. 2). A more trivial explanation is that Taq and Exo−Kl selected sequences with appropriately oriented A-T-rich flexible regions to promote binding, but this is argued against by the specific selection of the T7 and T3 core nucleotides. Since the starting pool had a 25-nucleotide random region, we calculate that only ∼1 in every 700 oligonucleotides would have contained a 7-nucleotide region corresponding to the T7 or T3 core region, let alone one that was in the appropriate position in the sequence to induce strong binding. This argues for the core promoter regions having specialized properties that promote binding beyond a high A-T content.
The phage core sequences in Taq1 and Exo−Kl, as well as the unrelated sequence in Taq2, were dependent on surrounding sequences to promote strong binding, and also highly orientation dependent (Tables 2, 3, and 5). Even a one-nucleotide shift in the position of the 3′ primer terminus dramatically weakened binding. If it is assumed that the 3′ terminus represents an anchor point for a primer-dependent polymerase, then moving the terminus by even a single nucleotide would change sequences that contact various domains of the polymerase. The observation suggests that the elements of the selected sequences that induce strong binding must be positioned precisely with specific polymerase binding domains. In this regard, binding is similar to the recognition of sites by sequence-specific DNA binding proteins such as promoter-dependent polymerase, enhancer, or restriction enzymes.
Sequence logos can be used to analyze the sequence conservation and diversity between a set of related nucleotide sequences (32). A logo comparing 17 T7 RNAP bacteriophage promoter binding sites is shown in Fig. 8 along with a logo comparing the sequences recovered from round 7 of PT-SELEX with Exo−Kl. The level of relative conservation of specific nucleotides between the two alignments is striking. It is notable that the 5′-ACTATAG-3′ sequence (nucleotides 12 to 18 in the Exo−Kl alignment and −6 to 0 in the T7 RNAP alignment) shared between the proteins showed almost identical conservation in bits. That is, in the region marked by bars at the bottom of the figure, the predominant sequence (reading the topmost letter—the consensus) is ACTATAG for both the Klenow SELEX and natural T7 RNA polymerase promoters.
Klenow sequences compared to T7 promoters using sequence logos. The aligned Klenow (Exo−Kl) sequences on the top correspond to those in Fig. 5. Below them are sequence logos of selected polymerase binding sites (23). A logo represents sequence conservation measured in bits at each position of the aligned sequences by the height of a stack of letters. Within each stack, the frequency of bases is shown by their relative heights. On the top of each logo stack is an “I” beam symbol that shows the likely variation of the stack height based on the number of sequences (35). To the right of each sequence (top panel) is the individual information for that sequence, in bits (35). The sequence logo for selected Klenow sequences is aligned with that of bacteriophage T7 promoters, and the common regions are marked with boxes. The open-stranded region for the DNA to which Klenow binds is shaded in light green in positions 16 to 18. The orientation of T7 polymerase on DNA has been determined (22, 41), as shown by the sine wave for which the peak represents the major groove facing the protein. Bases that exceed 1 bit are likely to represent non-B-form DNA (42) as confirmed experimentally by crystal structures (43–45). The region matches between Klenow (12 to 18) and the T7 promoter (−6 to 0) include the two possible flipping bases at −3 and −4 of the T7 promoter (circles). The last base of the common region corresponds exactly to the first base of transcription of the T7 promoter (triangle, base 0). The single-stranded region is just to the right of the potentially flipping bases and to the left of the initiation base. This correspondence suggests that the strong binding of Klenow to this sequence may be based on an opened DNA structure.
To test the resemblance quantitatively, we used the T7-like promoter model developed by Chen and Schneider (33). To build this model, 76 promoters from six T7-like bacteriophage had been combined into a single model, as shown in Fig. 4E of that paper. That model was successful in predicting T7-like promoters and led to the discovery of T7 islands, transposon-like genetic structures that apparently use T7-like promoters (34). Remarkably, the consensus of the 76-site T7-like promoter model from −6 to 0 is also 5′-ACTATAG-3′, the same as the Klenow SELEX consensus. To compare these precisely, the −6 to 0 individual information weight matrix model (35) for the 76-site model was scanned over the Klenow SELEX sequences. Every Klenow SELEX sequence had a positive information content by this evaluation, ranging from 0.2 to 12.6 bits. The second law of thermodynamics implies that positive values of individual information measured in bits imply functional binding sites (35–37), providing evidence in favor of the Klenow SELEX sequences being in the T7-like promoter class.
Sequence logos for Taq1 and T3 RNAP are shown in Fig. 9. Quantitatively, the region −7 to 0 of the 76-site individual information model matches the first 5 Taq SELEX sequences at 13 bits, while the other sequences are below zero bits, suggesting that they are a different class of molecules. In this case, the similarities are not as clear. This could be due in part to Taq polymerase selecting two diverse sequence motifs in the SELEX experiments (Fig. 2, groups 1 and 2). However, a trend between the shared sequence of Taq1 (5′-CACTAAAG-3′, nucleotides 9 to 16 for Taq and −7 to 0 for T3 RNAP) and the T3 promoter sequences is still evident, especially for the last 4 nucleotides. The similarities between the logo analyses with Klenow and Taq and phage promoters suggest that the uncommonly strong binding between the DNA polymerase and these specific sequences may be dictated by the same parameters that specify binding of the phage RNAPs to promoter sequences. The 76-site model closely resembles the common sequences for Taq and Klenow sequences. This result is consistent with the evolutionary connection between these polymerases.
Taq sequences compared to T3 promoters using sequence logos. The aligned Taq sequences on the top correspond to those in Fig. 2A. The conservation for the aligned sequences is shown as a logo in the middle. See Fig. 7 legend for details of the alignment and the logos. The logo on the bottom is for bacteriophage T3 promoter sequences (33). A large box shows the apparent alignment between the Taq-selected sequences and the T3 promoters (33). The alignment for T3 is like that for T7. The potentially flipping bases (above the sine wave [42]) are less clear because there are fewer sequences; four common bases are marked with circles. The triangle indicates the first base of the mRNA transcript. The wave on the T3 logo is in a dashed line to indicate that we are unaware of data that assign the binding—though it almost certainly has to be the same as for T7.
Exactly what properties of the promoter core sequence allow strong binding was not clear. Taq2, which binds even tighter than Taq1, had an A-T-rich region that did not match a phage RNAP core, but it was further upstream of the 3′ recessed terminus, and Taq2 bears little overall similarity to Taq1. This indicates that there are other sequences not related to phage promoters that can promote high-affinity binding, perhaps by a different mechanism. Further complicating the analysis was the inconsistent positioning of the core sequence in Taq1 and Exo−Kl-1. The T7 core region of Exo−Kl-1 was positioned over the 3′ recessed terminus with 4 nucleotides in the duplex region and 3 in the single-stranded 5′ overhang (Fig. 5). In contrast, in Taq1, the T3 core ended 5 nucleotides upstream of the 3′ recessed terminus (Fig. 2). Therefore, the core promoter regions of Taq1 and Exo−Kl-1 are shifted by several nucleotides. Although this might suggest that these sequences have different functions for the different enzymes, it is important to note that Exo−Kl was highly limited in the selection process by the apparent requirement to select a second BbsI site that could generate the longer overhangs required for Exo−Kl to gel shift the substrate (see Results and Fig. 5). In fact, the first 2 nucleotides of the selected T7 core sequence (AC) are part of the selected BbsI restriction site. Unlike the Taq1 sequence, Exo−Kl-1 would position the Klenow active site in nearly the same position relative to the core as the T7 RNAP active site is positioned in the T7-DNA crystal structure (22). It is unclear whether this positioning was forced by the prerequisite selection of the BbsI site or if there was a more complex interplay for selection of the T7 core sequence and BbsI site. Finally, it is also possible that phage core promoter sequences have special properties that would induce their selection by many polymerases, even non-family A and non-DNAP-like polymerases. Although this was not the case for viral RTs (2), more enzymes would have to be tested to draw any conclusions.
In addition to the biological implications discussed above, this work also demonstrated that the high-affinity Taq primers dramatically outcompeted other primers in PCRs, even when only a single high-affinity primer was included as the forward or reverse primer in a PCR see (Fig. S1 and S2 in the supplemental material). Primer bias in PCRs is known to occur and commonly results from differential hybridization kinetics of primers to target DNA (14–16) but may also be related to DNA polymerase primer specificity (38, 39). This phenomenon can complicate genome sequencing and other quantitative protocols requiring PCRs (e.g., multiplexing, transcriptome sequencing [RNA-seq], DNA sequencing [DNA-seq]), where short nonspecific primers (e.g., random hexamers) are often used in the PCR step. The extreme primer bias demonstrated in the current work was not related to the thermostability of the primers used but correlated with more-stable binding of Taq to the SELEX-selected primers (Table 4) and would therefore be consistent with the DNA polymerase primer specificity noted above. Since PT-SELEX is designed to select only the primers that bind with the highest affinity, it does not reveal information about primers that may bind preferentially but to a lesser extent, and it is not clear how such primers would affect PCRs. There may be a spectrum of different Taq binding affinities for primers, or the selected primers may represent rare sequences that bind with uncommonly high affinity. If the latter were the case, DNA polymerase primer specificity might not be a major issue in most multiplex reactions with multiple primers. In contrast, if a spectrum of different primer affinities exists, this could complicate quantitative analysis in multiplexing and other PCR protocols. In the future, we plan to use PT-SELEX to determine if other thermostable polymerases also show strong sequence bias and if the recovered sequences are related to those found for Taq.
MATERIALS AND METHODS
Materials.A 3′→5′ exonuclease− mutant of Klenow polymerase (referred to as Exo−Kl), containing the mutations D355A and E357A, was purchased from New England BioLabs, as were wild-type Klenow, Taq DNA polymerase, BbsI, and T4 polynucleotide kinase (T4 PNK). dNTPs were purchased from Roche Applied Sciences. Radiolabeled [γ-32P]ATP was from PerkinElmer. Sephadex G-25 spin columns were from Harvard Apparatus. Oligonucleotides were from Integrated DNA Technologies. The PCR blunt end cloning kit was from Agilent. The Miniprep DNA preparation kit was from Qiagen. Nitrocellulose filter disks (25 μm, 0.45-μm pore size, Protran BA 85) were from Whatman. The LigaFast rapid ligation kit was from Promega. All other reagents were obtained from Thermo Fisher Scientific, Inc., Sigma-Aldrich Co., or VWR. Graphs were produced and analyzed using SigmaPlot. The sequence logos and aligned sequences in Fig. 8 and 9 were generated using makelogo 9.59, alist 6.63, and alo 1.12, which are available at https://alum.mit.edu/www/toms (32).
5′-32P end labeling of DNA oligonucleotides.Twenty-five picomoles of the 41-nucleotide primer strand oligonucleotide or 500 pmol of primer 1 (see below) was 5′-32P end labeled using T4 PNK. The labeling reaction was performed at 37°C for 30 min, in accordance with the manufacturer's protocol. Reaction mixtures were shifted to 70°C for 15 min to heat inactivate the PNK. The DNA was then centrifuged on a Sephadex G-25 column to remove any excess radiolabeled nucleotide.
Preparation of DNA-DNA duplexes.Duplexes were prepared by mixing 2 pmol of 41-nucleotide 5′-32P-end-labeled DNA from above and 2 pmol of 45-nucleotide template (see Results for a full list of sequences) in 15 μl of buffer containing 50 mM Tris-HCl (pH 8.0), 80 mM KCl, 1 mM dithiothreitol (DTT), and 0.1 mM EDTA. Reaction mixtures were placed at 80°C for 5 min and then allowed to slowly cool to room temperature prior to use.
Preparation of starting material for PT-SELEX.Approximately 400 pmol each of radiolabeled primer (5′-GCCTGCAGGTCGACTCTAGA-3′ [primer 1]) and unlabeled template [5′-GCATGAATTCCCGAAGACGC(N)25TCTAGAGTCGACCTGCAGGC-3′, where N is any base] was mixed in a 50-μl volume containing 50 mM Tris-HCl (pH 8), 1 mM DTT, and 80 mM KCl. The material was heated to 65°C for 5 min and then slow cooled to room temperature to form hybrids. The hybrid reaction mixture was diluted into a total volume of 100 μl containing 0.5 mM dNTPs, 6 mM MgCl2, and 3 U Klenow polymerase and incubated for 45 min at 37°C. The product was purified by running on a 12% native polyacrylamide gel in Tris-borate EDTA (TBE) (40) and cutting out bands corresponding to the double-stranded hybrid. The hybrid was eluted from the gel overnight in 500 μl of 25 mM Tris-HCl (pH 8) at 4°C and then passed through a 0.45-μm polyethersulfone membrane syringe filter. The material was precipitated at −20°C in 2 volumes of 100% ethanol and 1/10 volume of 3 M sodium acetate (pH 7) (final volume, ∼1,500 μl) and then spun for 30 min at ∼15,000 × g to pellet the DNA. The pellet was washed with 500 μl of 70% ethanol and then dried in a Savant DNA120 speed vacuum. Collected hybrids were incubated with 20 U of BbsI at 37°C for 3 h in a final volume of 100 μl using the buffer supplied by the manufacturer. Digested products were separated from any uncut material by electrophoresis on a 12% native polyacrylamide gel and recovered as described above. Each reaction volume typically yielded about 50 to 100 pmol of final product.
PT-SELEX with Exo−Kl and Taq polymerase.An overview of the PT-SELEX process is presented in Fig. 1. For the initial selection, ∼200 pmol of cleaved hybrid was incubated with 20 pmol Exo−Kl or Taq polymerase for 1 h in 100 μl of 50 mM Tris-HCl (pH 7), 50 mM KCl, 1 mM DTT, 10 mM MgCl2, and 50 μg/ml bovine serum albumin (BSA) (Klenow buffer) or 10 mM Tris-HCl, 50 mM KCl, and 1.5 mM MgCl2 (Taq polymerase buffer) at room temperature. After addition of 6× gel loading buffer, consisting of 30% (vol/vol) glycerol, 0.25% (wt/vol) bromophenol blue, and 0.25% (wt/vol) xylene cyanol FF, samples were run on a 7% native gel to separate bound and unbound materials. Bound material shifted toward the top of the gel and was recovered as described above.
After selection, the hybrids' recessed ends were filled in using Klenow polymerase in reaction mixtures containing 0.2 U of enzyme, 50 mM Tris-HCl (pH 8), 100 μM dNTPs, 6 mM MgCl2, 80 mM KCl, and 1 mM DTT, for a total volume of 50 μl. After 10 min at 37°C, reaction mixtures were phenol extracted and precipitated with ethanol in the presence of 50 μg of glycogen. Following this, the recovered material was ligated for 20 min to a 5- to 10-fold excess of the hybrid duplex of 5′-ATAGCATGAATTCGCAGAAGACCC-3′ and 5′-GGGTCTTCTGCGAATTCATGC-3′ (no phosphate on 5′ ends), using the LigaFast rapid ligation kit. For the first round, the ligation was performed in a volume of 30 μl; subsequent rounds were performed in 10 μl as per the manufacturer's protocol. Note that the duplex can potentially ligate to either end of the selected material; however, the vast majority of ligations will be to the end with random nucleotides (Fig. 1), as the other 5′ end is mostly nonphosphorylated. Only those primers phosphorylated during the radiolabeling step would be substrates for ligation, and these represent just a few percentages of the total primers used. Even if ligation to this end does occur, the resulting PCR products would be eliminated based on their size in the PCR step below.
The entire ligation mixture from above was then PCR amplified by Taq polymerase in a 400-μl reaction mixture containing 400 pmol each of the primers 5′-GCCTGCAGGTCGACTCTAGA-3′ (32P end labeled) and 5′-GCATGAATTCGCAGAAGACCC-3′, in Taq buffer. The reaction mixtures were divided equally into 4 tubes and PCR amplified for 8 to 14 cycles of 94°C (30 s), 50°C (30 s), and 72°C (30 s), followed by a final 5-min extension at 72°C. The number of amplification cycles was controlled to prevent overamplification, which results in a smeared rather than a discrete product on a nondenaturing gel. Typically, aliquots were removed from the PCR mixtures after 8, 10, 12, or 14 cycles, added to 6× gel loading buffer, and then run on a 12% nondenaturing gel as described above. The correct-size products were excised and recovered as described above. In some rounds, the recovered products were used to make more PCR product (∼0.1 pmol of recovered product was amplified for 8 to 10 cycles) in order to ensure a yield of at least 25 pmol for the next SELEX round. Recovered material was combined, BbsI digested, purified as described above, and then subjected to the next round of SELEX. This selection process was repeated 6 times for both Taq polymerase and Exo−Kl, with subsequent rounds using ∼1/10 (rounds 2 to 4) or 1/20 (rounds 5 to 7) the enzyme (mole/mole) to the recovered material, until no increase in binding was detected between rounds. Also, only one-half of the material from the ligation mixture (see above) was used in PCRs after round 1. After round 2, 1/5 of the material was also saved from each round for use in Kd determinations. After reaching round 7, PCR material from each selection pool was inserted into a vector and cloned into bacteria using the Strataclone Blunt Ended PCR kit from Agilent Technologies according to the manufacturer's instructions. Isolated colonies were grown overnight in 3 ml of 50 μg/ml ampicillin-LB broth. Insert-containing plasmids were purified using a Qiagen Spin Miniprep kit and sequenced.
Determination of the equilibrium dissociation constant by gel shift assay.To determine Kd by gel shift assay, oligonucleotides based on the recovered sequences from PT-SELEX were purchased, 5′ end labeled on the primer strand (36 nucleotides for Exo−Kl and 41 nucleotides for Taq unless otherwise noted), and hybridized to the template strand (45 nucleotides unless otherwise noted) as described above. Hybrids were purified from 12% native gels as described above. These were mixed with various amounts of either Taq polymerase or Exo−Kl at a final concentration of 2 nM in either Klenow or Taq polymerase buffer (see above). The enzyme concentrations were 0, 1.1, 3.1, 6.3, 12.5, 25, and 50 nM for Taq polymerase and 0, 25, 50, 100, 200, 400, and 800 nM for Exo−Kl, unless otherwise indicated. The enzymes and hybrids were incubated for 5 min at room temperature in a final volume of 15 μl followed by addition of 6× gel loading buffer. Samples were then run at 110 V on a 7% polyacrylamide native gel. The products were visualized on a FLA-7000 phosphorimager (Fujifilm), and the ratio of amount of product shifted to amount unshifted was quantified with Multi Gauge software. Kd was determined by plotting the ratio of material shifted ([shifted]/[shifted + unshifted]) to the concentration of the enzyme and fitting the data by nonlinear least-square fit to the following quadratic equation: [ED] = 0.5 ([E]t + [D]t + Kd) − 0.5 {([E]t + [D]t + Kd)2 − (4 [E]t[D]t)1/2}, where [E]t is the total enzyme concentration and [D]t is the total primer-template concentration. As these experiments used relatively large amounts of nucleic acid starting material (2 nM) and the conditions were not optimized for binding to the polymerases, the gels were used only to compare binding to different constructs. These “apparent” equilibrium dissociation constants (referred to as Kd,app,gelshift) were determined for comparisons between the various constructs.
Determination of Kd, koff, and t1/2 for binding of Taq to primer-templates at 23°C (room temperature) and 60°C using nitrocellulose filter binding.For dissociation constant (koff) determinations, reaction mixtures were in Taq polymerase buffer (see above) containing 5 nM gel-purified radiolabeled primer-template construct and 5 nM Taq polymerase in a volume of 90 μl. To start the analysis, 500 nM (final concentration in reaction mixtures) of unlabeled primer-template was added in a volume of 10 μl in Taq buffer. Ten-microliter aliquots were removed at different time points (depending on how long the construct bound to Taq) and immediately applied to nitrocellulose filter disks under vacuum. Each disk was washed twice with 1 ml of wash buffer (25 mM Tris-HCl [pH 7.5], 10 mM KCl), air dried, and counted in a scintillation counter. A graph of bound counts versus time was constructed, and the koff value was calculated by fitting the curve to an equation for two-parameter exponential decay (y = ae−bx, where a is the value for bound substrate at time zero and b is the koff value) using SigmaPlot. Averages from 3 or more independent experiments (±standard deviations [SD]) were used to calculate koff. The half-life (t1/2) for binding was calculated from koff using the equation t1/2 = 0.693/koff. Kd determinations used the same buffer conditions as those described above with the addition of 0.1 mg/ml BSA. The reactions were performed in 1 ml of buffer that contained 2 pM gel-purified radiolabeled primer-template construct. Different amounts of Taq polymerase were added based on pilot experiments to approximate the Kd. Material was incubated at 23°C for 10 min and then applied to nitrocellulose filters as described above. The Kd was calculated using the equation described above in “Determination of the equilibrium dissociation constant by gel shift assay.” Note that this approach could not be used to accurately assess Kd values at 60°C due to temperature fluctuations that occur during application of the sample to the filter and washing of the filter.
Preparation of constructs for competition PCR experiments.Constructs were prepared by PCR amplification using 50 ng of plasmid pNL43 and 50 pmol of each primer for 25 cycles of 94°C (30s), 50°C (1 min), and 72°C (1 min), followed by 1 cycle for 4 min at 72°C. Samples were electrophoresed on a 6% native polyacrylamide gel, located by UV shadowing, and purified as described above. The primers used for the reactions are described in Table S1 in the supplemental material.
Competition PCR experiments with Taq polymerase.Reactions were performed in standard Taq polymerase buffer containing 200 μM dNTPs and 2.5 units of Taq polymerase in a 100-μl final volume. Reaction mixtures contained a 448-nucleotide template construct (prepared as described above) containing a 400-nucleotide region of the HIV genome (nucleotides 336 to 735 in the HIV LAI virus sequence) that was flanked by two 24-nucleotide regions at the 5′ and 3′ ends (see Fig. S1A and S2A in the supplemental material). The 24-nucleotide regions contained 20 nucleotides that could bind to different sets of primers followed by a fixed 4-nucleotide region derived from the 5′ overhang sequences of Taq1 (5′ end of construct) or Taq2 (3′ end of construct) (Table 1). Reaction mixtures included 1 ng of template and 50 pmol of each primer, including one radiolabeled primer (5′ end primer, in each case denoted with an asterisk in Fig. S1 and S2). For competition reactions, a second primer-template set was included at the same concentration without radiolabel. Cycling conditions were 94°C (30 s), 53°C (30 s), and 72°C (30 s). Ten-microliter aliquots were removed at 10, 13, 16, 19, 22, 25, and 28 cycles and added to 2 μl of 6× agarose gel loading buffer. Samples were electrophoresed on a 1% agarose gel, fixed in 10% trichloroacetic acid (TCA), and dried onto filter paper as described previously (40). Dried gels were analyzed using a phosphorimager.
ACKNOWLEDGMENTS
This work was supported in part by a grant from the National Institute of General Medical Sciences (GM116645) awarded to J.J.D. and in part by funds to T.D.S. from the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research.
FOOTNOTES
- Received 27 September 2017.
- Accepted 11 January 2018.
- Accepted manuscript posted online 16 January 2018.
- Address correspondence to Jeffrey J. DeStefano, jdestefa{at}umd.edu.
↵* Present address: Vasudevan Achuthan, Dana-Farber Cancer Institute, Harvard Medical School, Boston, Massachusetts, USA.
Citation Fenstermacher KJ, Achuthan V, Schneider TD, Destefano JJ. 2018. An evolutionary/biochemical connection between promoter- and primer-dependent polymerases revealed by systematic evolution of ligands by exponential enrichment. J Bacteriol 200:e00579-17. https://doi.org/10.1128/JB.00579-17.
Supplemental material for this article may be found at https://doi.org/10.1128/JB.00579-17.
REFERENCES
- Copyright © 2018 American Society for Microbiology.
