CODEHOP-mediated PCR – A powerful technique for the identification and characterization of viral genomes

General CODEHOP Design and PCR Strategy

CODEHOPs are derived from highly conserved amino acid sequence motifs present in multiple alignments of related proteins from a targeted gene family. Each CODEHOP consists of a pool of primers where each primer contains one of the possible coding sequences across a 3–4 amino acid motif at the 3' end (degenerate core) (Figure 1A) [2]. Each primer also contains a longer sequence derived from a consensus of the possible coding sequences 5' to the core motif (consensus clamp). Thus, each primer has a different 3' sequence coding for the amino acid motif and the same 5' consensus sequence. Hybridization of the 3' degenerate core with the target DNA template is stabilized by the 5' consensus clamp during the initial PCR amplification reaction (Figure 1B). Hybridization of primers to PCR products during subsequent amplification cycles is driven by interactions through the 5' consensus clamp.

Conserved amino acid motifs used for CODEHOP design are identified by alignment of related proteins from a targeted gene family using computer programs such as the Clustal W multiple alignment program [12]. Optimal blocks contain 3–4 highly conserved amino acids with restricted codon multiplicity from which the 3' degenerate core is derived; the presence of serines, arginines and leucines are not favored due to the presence of six possible codons for each amino acid. In addition, optimal blocks contain 5 or more conserved amino acids from which the 5' consensus clamp is derived. These blocks of conserved amino acid sequences should be situated in close enough proximity to allow efficient PCR amplification between blocks yet distant enough to flank a region of significant sequence information.

CODEHOP PCR Amplification, Product Cloning and Sequence Analysis

CODEHOP PCR amplification has been performed using classical and touch-down approaches with a hot-start initiation [2]. More recently, thermal gradient PCR amplification has been used to empirically determine optimal annealing and amplification conditions for the pool of primers [11]. Different buffers, salt concentrations, and enzymes have been employed with varying success due to differences in DNA template preparation and the unknown nature of the targeted sequence. PCR products are either sequenced directly or after TA-cloning.

In this review, sequences were compared by BLAST analysis [14] and multiple alignment using Clustal W [12]. Phylogenetic analysis of the multiply aligned sequences was performed using protein distance and neighbor-joining analysis implemented in the Phylip analysis package [15]. Bootstrap analysis was also performed with 100 replicates and a consensus phylogenetic tree was determined. For the phylogenetic analysis, positions in the multiple alignment containing gaps due to insertions or deletions within the sequence blocks were eliminated.

Limiting degeneracy to increase sensitivity

While the "TVG-IYG" herpesvirus assay demonstrated the ability to detect disparate herpesvirus species in high titer virus cultures in vitro, the detection of limiting amounts of virus in tissue samples in vivo was problematic. This was especially true in sections obtained from formalin-fixed, paraffin-embedded tissue blocks which contained small amounts of degraded DNA. The degeneracy of the primer pool, ie. the number of different primers necessary to encode all codon possibilities for the specified block of conserved amino acids, plays a direct role in the sensitivity of the PCR amplification. Whereas highly degenerate primers consisting of pools of hundreds or thousands of primers with different DNA sequences may allow amplification of DNA templates present in high copy number, as found in cultured virus stocks, they are less successful in amplifying low copy numbers of DNA templates found in virus infected tissues in vivo, especially in formalin-fixed tissue. As the degeneracy increases, the concentration of the primer or primers that will participate in the desired amplification reaction decreases and can become suboptimal. Conversely, the vast excess of primers not participating in the amplification of the targeted gene can cause non-specific amplification which can, in turn, inhibit or mask the amplification of the desired target.

As indicated in Table 1, the degeneracy of the primers utilized in the "TVG-IYG" assay ranged from 48–1024. This level of degeneracy was driven by the number of nucleotide possibilities encoding the targeted amino acids at each position as well as by the number of amino acid positions allowed to be degenerate. Figure 5A shows the DFA/DFAS/QAHN sequence block produced by Block Maker from multiple alignments of 11 different herpesvirus polymerase sequences. Figure 5C shows the consensus amino acids at each position, as determined by the CODEHOP algorithm, which are boxed and bolded with the alternate amino acids positioned above. The original primer manually derived from this motif, "DFA" is, in fact, completely degenerate, with multiple codons provided for each amino acid position, except the ultimate proline (P) residue, yielding a pool of 512 different primers [8]. Because the performance of this primer was consistently suboptimal in samples with limiting template, the overall structure and degeneracy of the primer was altered by designing a PCR primer "DFASA" from the same sequence motif using the CODEHOP methodology. This primer had an 11 bp 5' consensus region and a 3' degenerate core containing multiple codons at 5 amino acid positions resulting in a pool of 256 different primers (Figure 5C). The "DFASA" primer was successfully used to amplify extremely low amounts of viral DNA in a background of genomic DNA from paraffin-embedded formalin-fixed tissue in the discovery of the macaque homolog of Kaposi's sarcoma-associated herpesvirus, called retroperitoneal fibromatosis herpesvirus (RFHV) [9]. Subsequent estimates of virus copy number using real-time quantitative PCR indicated a level of RFHV DNA in the available samples that was 1/100–1/1000 of a single copy cellular gene (unpublished observations). The "DFASA" primer has been successfully used to identify a number of novel alpha-, beta- and gammaherpesviruses in a wide variety of host organisms (see Additional File 1: "DFASA-GDTD1B assay").

Due to the presence of a highly conserved leucine (L) at block position 7 within the "DFAS" motif (Figure 5) which significantly increased the degeneracy of the primer pool with its six possible codons, an additional CODEHOP was designed from the "QAHN" motif immediately downstream of "DFAS" to further decrease degeneracy. The "QAHNA" primer had an 11 bp 5'consensus region and a 3' degenerate core containing multiple codons at 4 amino acid positions resulting in a pool of 48 different primers (Figure 5C). This CODEHOP has been successfully used to identify several primate rhadinoviruses related to KSHV in tissue samples with limiting amount of viral DNA [10, 19], see also Additional File 1.

Primer bias and specificity

The primers developed for the "TGV-IYG" assay were designed to amplify polymerase fragments from herpesviruses of all three subfamilies based on conserved motifs within the known sequences. However, very few sequence motifs were absolutely conserved between the most divergent herpesviruses. For example, the catfish ictalurid herpesvirus (IHV) lacked the "KGV" motif from which the initial "KGV" primer was derived (Figure 6). Furthermore, numerous sequence differences were present in the IHV DNA polymerase within the DFAS/QAHN motif which was otherwise highly conserved in other herpesvirus species (highlighted residues in Fig. 5B). Because of these differences, the IHV sequence was excluded from the primer design of the "DFA", "DFASA" and "QAHNA" PCR primers. As shown in Figure 5C, the "DFA" and "DFASA" primers have mismatches with the IHV sequence at the alanine (A) and leucine (L) codons (Block positions 5 and 7, respectively; Figure 5B) and the "QAHNA" primer mismatches at three codon positions (Block positions 13–15; Figure 5B), all within the 3' degenerate cores. Figure 8 shows the presence of nucleotide mismatches with the IHV sequence throughout the different primers (black highlighting). Thus, the lack of the "KGV" motif and sequence differences in the "DFA" primer strongly biased the "TGV-IYG" assay against IHV-like herpesvirus sequences. In order to identify IHV-like herpesviruses, new primers would have to incorporate these sequence differences.

The "DFA" and "DFASA" primer pools were originally designed using only the alanine (A) codon at block position 5 in the "DFAS" motif and did not include the glutamine (Q) codon found in that position of the motif in HHV6 and HHV7, "DFQS" (highlighted, Figure 5A, B). The nucleotide mismatches in this region are shown in Figure 8. While the "DFA" and "DFASA" primers are biased by design against HHV6 and HHV7, they have been used successfully to detect betaherpesviruses related to HHV6 and HHV7 [8]. This suggests that mismatches 13–14 nucleotides from the 3' end of the primer, do not have major affects on the utility of the primers, especially when viral template is not limiting.

More significant bias against HHV6- and HHV7-like herpesviruses was present in the "TGV" primer used in conjunction with the "IYG" primer in the secondary nested PCR reaction in the "TGV-IYG" assay (see Figure 2B). The "TGV" primer contains the partial valine (V) codon "GT" at its 3' end (Block position 11; Figure 3C). Since both HHV6 and HHV7 contain alanine (A) (codon = GCN) at this position (highlighted in Fig. 3A, B), the "TGV" primer would mismatch at the 3' terminal nucleotide with both HHV6- and HHV7-like sequences. This mismatch occurs at the 3' end of the "TGV" primer and is predicted to significantly impair polymerase extension. To remove this bias, the "TGV" primer was redesigned as the "VYGA" primer removing the 3' terminal "GT" of the valine codon and the terminal degenerate position of the glycine (G) codon. The "TGV" primer contained an additional bias against amplification of HHV6-like sequences due to the use of only the phenylalanine (F) codons (TTY) (Block position 8) at a position encoding valine (V) in both HHV6 and HHV7 (highlighted in Figure 3A and 3B). To remove this bias, "VYGA" was designed to include both the valine (V) and (F) codons at this position. The total degeneracy of the "TGV" and "VYGA" primer pools remained the same, with 256 different primers, due to the loss of the degenerate codon position in the glycine, block position 10 in "TGV" and the gain of the degenerate codon positions in the valine, block position 8 in "VYGA".

The subsequent cloning and sequence analysis of new herpesvirus DNA polymerases from the rhadinoviruses, rhesus rhadinovirus (RRV) and alcelaphine herpesvirus 1 (AlHV1) [20, 21], revealed mismatches with the downstream "IYG" primer of the "TVG-IYG" herpesvirus assay. The "IYG" primer (a reverse orientation primer) includes the codons (ATH) for isoleucine (I) at its 3' end (Block position 1; Figure 4C). Both RRV and AH1 contain a valine (V) codon (GTN) at this position (highlighted in Figure 4A). Thus, "IYG" is biased against RRV-like or AH1-like rhadinoviruses due to a T-C mismatch at the 3' end of the primer. To eliminate this bias, the "IYG" primer was redesigned as "GDTD1B" to remove the isoleucine position within the 3' degenerate core (Figure 4C) and, in addition, the length of the 5' consensus clamp was increased.

Decrease in size of the amplification products

Because typical tissue samples especially paraffin-embedded formalin-fixed tissue contain degraded DNA with sizes averaging near 300–500 bp in length, we decided to decrease the maximal amplification product size of the herpesvirus assay. The initial amplification product of the "TGV-IYG" assay (DFA-KG1) was ~800 bp (Fig. 2B). To reduce the initial amplification product size, a hemi-nested PCR assay was developed in which the newly designed downstream anti-sense primer "GDTD1B" targeting the highly conserved "YGDT" motif was used in a primary PCR amplification with the new upstream primer "DFASA". This amplification yields an approximate 500 bp PCR product (Figure 2B). This initial PCR product is then used as template in a secondary PCR amplification using the nested primer "VYGA" with the downstream anti-sense primer "GDTD1B". This amplification yields a PCR product of approximately 200 bp (see Figure 2B). These modifications produce amplification products close to the average size of degraded DNA present in fixed tissue.

Using the web-based software to design CODEHOP PCR primers to a conserved viral gene

The amino acid sequences of the TS genes from five herpesviruses were obtained using BLAST analysis of the NCBI protein database with the KSHV TS sequence as probe. The TS sequences from KSHV, VZV, EHV2, HVS and AtHV3 (Figure 16) were provided as input to ClustalW [28] and a multiple alignment was obtained. As shown in Figure 13, several regions of highly conserved sequences were present in the TS sequence alignment, and the positions of the "RHFG" and "DMGL" motifs targeted above are indicated. In order to predict CODEHOP PCR primers, the sequences of the TS genes were provided as input to the BlockMaker program of the Blocks Database [4] and a series of conserved sequence blocks were identified (ex., Gibbs Blocks, Figure 17). Alternatively, the ClustalW alignment, itself, could be provided as input to the "Multiple alignment processor" of the Blocks Database [29]. In order to compare a computer-predicted CODEHOP with the manually derived CODEHOP (DMGLB), the TS Block_E containing the "DMGL" motif (Figure 17) was directly input to the CODEHOP program [3] using all default values except that the consensus region was elongated by increasing the temperature setting from the default 60°C to 70°C. The primers predicted from the complement of Block_E were examined in order to obtain a primer from the complementary strand which could be used in conjunction with the upstream TS primer RHFGA, described above. The underlined primer targeting the "DMGL" motif was chosen and named DMGLXB (Figure 18) and was compared with the manually designed DMGLB primer in Figure 14. Whereas "DMGLB" was purposefully biased by using the KSHV sequences in the 5' consensus clamp, the "DMGLXB" is "unbiased" in design with the 5' consensus sequence derived from the most frequently used codons in the human genome. The DMGLXB sequence was examined for potential stem loop structures that could compromise the function of the primer. As shown in Figure 14, a putative stem-loop structure was identified which is indicated by the underlined nucleotides in Figure 14B and 14C. To destablize this structure, the proline codon within the "DMGLGVP" motif was changed from the computer predicted "CCC", the most frequently used codon in humans, to "CCA", another common human codon, as shown in Figure 14. This yielded a revised CODEHOP, called "DMGLX1B" (shown as the complementary sequence in Figures 14B and 14C), in which the stem-loop structure was destabilized by substituting an A for the highlighted C in Figure 14C. The DMGLX1B antisense primer could then be used in combination with the RHFGA sense primer to amplify unknown TS genes.

Other examples of CODEHOP PCR primers designed from multiple alignments of the herpesvirus DNA polymerase sequences using the Web-based CODEHOP software are shown in Figures 3, 4, 5, 6. The VYG1A primer designed from the conserved VYG motif shown in Figure 3 is aligned with the original manually designed "TGV" and "VYGA" primers. The computer-predicted "YGDTB" primer designed from the conserved GDTD motif is aligned with the original "IYG" and "GDTD1B" primers (Figure 4). In the prediction of this primer, the conserved sequence block identified by BlockMaker from the sequences shown in Figure 4A, extended only from amino acid position 1 – 10, which was the limit of the conserved sequence block determined by BlockMaker. The CODEHOP software indicated the necessity to add additional nucleotides to the 5' end of the "YGDTB" primer to obtain the minimal length for the 5' consensus region of the primer. As such, the amino acid sequences of block positions 11–13 were obtained manually and compared in order to derive the eight terminal nucleotides for "YGDTB" (overlined in Figure 4C).