|Home | About | Journals | Submit | Contact Us | Français|
Previous relatively small studies have associated particular amino acid replacements and deletions in the HIV-1 nef gene with differences in the rate of HIV disease progression. We tested more rigorously whether particular nef amino acid differences and deletions are associated with HIV disease progression. Amino acid replacements and deletions in patients' consensus sequences were investigated for 153 progressor (P), 615 long-term nonprogressor (LTNP), and 2,311 unknown progressor sequences from 582 subtype B HIV-infected patients. LTNPs had more defective nefs (interrupted by frameshifts or stop codons), but on a per-patient basis there was no excess of LTNP patients with one or more defective nef sequences compared to the Ps (P = 0.47). The high frequency of amino acid replacement at residues S8, V10, I11, A15, V85, V133, N157, S163, V168, D174, R178, E182, and R188 in LTNPs was also seen in permuted datasets, implying that these are simply rapidly evolving residues. Permutation testing revealed that residues showing the greatest excess over expectation (A15, V85, N157, S163, V168, D174, R178, and R188) were not significant (P = 0.77). Exploratory analysis suggested a hypothetical excess of frameshifting in the regions 9SVIG and 118QGYF among LTNPs. The regions V10 and 152KVEEA of nef were commonly deleted in LTNPs. However, permutation testing indicated that none of the regions displayed significantly excessive deletion in LTNPs. In conclusion, meta-analysis of HIV-1 nef sequences provides no clear evidence of whether defective nef sequences or particular regions of the protein play a significant role in disease progression.
HIV-infected people can be categorized according to the number of years in which they progress to AIDS. Long-term nonprogressors (LTNPs) do not progress to AIDS even after more than 10 years of infection, and they maintain stable CD4 lymphocyte counts (5, 8). Nonprogression status may reflect differences in either in the host, in viral genetics, or in environmental factors. Within the virus, R77Q, a mutation in the HIV-1 vpr gene, was associated with both LTNP infection and impaired induction of apoptosis (38). However, this mutation was not statistically significant, and no other clearly attenuating mutations or deletions were detected (20). Most attention, however, has focused on role of the viral nef protein.
In rhesus monkeys infected with simian immunodeficiency virus (SIV), a model for studying AIDS pathogenesis (37), animals infected with nef-deficient SIV showed an attenuated course of infection (17, 30, 31, 51). nef was also a major determinant of pathogenicity in transgenic mice with AIDS-like symptoms induced by HIV-1 (27). Some patients with LTNP strains of HIV were found to have gross deletions in the nef gene (16, 33, 49), suggesting the importance of nef for HIV-1 progression in humans.
Previous studies related to phylogenetic analysis have reported that nef sequences from patients with different rates of progression do not form distinct clusters (28, 29, 40, 43). Each patient had sequences that clustered together and could be differentiated from those of the other patients, supporting the monophyletic origin of the infections. The absence of intragroup clustering suggested that no correlation existed between the phylogenetic relationship of the nef sequences and the progression rate in the patients (10). The differences in genetic distance between LTNP and progressors (Ps) were not statistically significant, suggesting that the degree of sequence variation in nef is unlikely to reflect the stage of HIV-1 disease (4).
Amino acids 25 to 36 in HIV-1 nef are important both for several well-defined in vitro functions of nef and for the pathogenicity of HIV-1 in humans, and nef's ability to enhance virion infectivity was fully restored when the deletion was repaired by the insertion of that region (8). Nef proteins derived from LTNPs and slow progressors (SPs) were found to be defective or far less capable of enhancing viral replication and/or viral infectivity in herpesvirus saimiri-transformed human T cells and peripheral blood mononuclear cells (PBMC) (24). The sizes of the deletions in the nef/LTR (long terminal repeat) region increased progressively from 84 to 1,400 bp during the 5-year follow-up period in one case of a SP (35). Gross defects were also present in the RNA-derived sequences of an LTNP individual because of a frameshift and the premature termination of the protein (4). HIV-1 sequences from the isolates or patient PBMC had similar deletions in the nef gene and in the region of overlap of nef and the U3 region of the LTR (16). There was a 36-bp deletion close to the 5′ end of nef that impaired nef function in an LTNP (8).
Many studies not only have described nef as carrying large deletions in LTNPs (16, 33) but also found a higher proportion of disrupted nef gene sequences in LTNPs. A study of six HIV patients who reached at least 11 years of age without or with mild symptoms revealed that LTNPs had higher proportions of disrupted nef sequences (10). Seven LTNPs, all belonging to the same cohort of infected hemophiliacs, had more defective nef sequences than in progressors; the number of disrupted nef sequences within each individual was significantly higher in LTNPs than in progressors (4).
The nef amino acid sequence has been reported to be highly polymorphic even within a particular subtype (4, 22, 28, 29, 40, 42, 53). Single amino acid deletions have been found predominantly at three locations that are structurally less defined loop regions: positions 8 to 11, 49 to 51, and 155 to 162 (25). Five variants (T15, N51, H102, L170, and E182) have been noted among LTNPs, whereas nine variants (N-terminal PxxP motif; A15, R39, T51, T157, C163, N169, Q170, and M182) have been noted among progressors (32). nef has been often changed at residues localized in the folded core domain at cytotoxic-T-lymphocyte epitopes (E105, K106, E110, Y132, K164, and R200); moreover, LTNP-associated variations occur in the core domain of nef. Recently, nef sequence variations have been found in the WL motif of the CD4 binding site, as well as a premature stop codon in infected LTNPs that could potentially contribute to the attenuation of the virus; however, these deletions were found to be insignificant (13).
There has been a broad agreement that grossly defective nefs are associated with an attenuated course of infection (17, 30, 31, 51) but rare in HIV-1 infection (32). Grossly defective nef genes or significant changes from relevant clade reference sequences were not identified in a study of 32 LTNP children (13). One study noted that the proportion of disrupted nef sequences within each patient was significantly higher in LTNPs compared to Ps; however, the proportions of individuals with nef defects (in LTNPs, 5 of 7, and in Ps, 6 of 8) were similar (4). No major defects have been reported in a few other studies (28, 39, 40). Another study of a small number of patients does not indicate that gross deletions play any major role in delaying or halting disease progression in infected drug abusers in Italy (11), and premature stop codons were observed at equivalent, yet low, frequencies among the different clinical groups (41). In addition, disease progression has been reported in a HIV patient with a virus grossly deleted of nef (26).
Thus, overall, most of these studies were based on observation or case study rather than systematic scientific evaluation (11). The objective of the present study was to determine in a substantially larger sample than investigated to date whether there is any association between disease progression and particular nef amino acid differences or deletions.
HIV-1 subtype B nef nucleotide and protein sequences from Pa, LTNPs, and UPs (those for whom the “progression” field was null) were collected from the Los Alamos National Laboratory (LANL) HIV Sequence Compendium 2009 (36) and the LANL website (http://www.hiv.lanl.gov) in November 2009. Details of all of the patients such as patient identification (ID) data and sampling details were also retrieved from the LANL website. Some of the information was retrieved from NCBI GenBank files and, in a few cases, from the LANL publications. A few of these sequences belong to previous studies (described in Table Table1).1). We used the LTNP and P definitions of the LANL database, which records the rate of progression of the patient as recorded by the study. Table S1 in the supplemental material provides details such as the patient IDs, country, number of sequences (defective and nondefective), mean lengths of genes and proteins, PubMed IDs, risk factor(s), year of infection, number of sampling time points, name of the isolate or clone, sampling year, patient age, health, viral load, CD4 count, number of days from infection, etc. (if available), for all LTNP and P patients considered in these analyses. The numbers of sequences and the numbers of patients in the present study are summarized in Table Table2.2. The nef sequence from the HIV-1 NL4.3 strain (206 amino acids) was used as a reference sequence for the whole analysis since nef is defective in the standard reference strain HXB2, containing a mutation at position 124, which produces a stop codon. NL4.3 has been used as a reference strain in previous studies (see, for example, references 15, 34, and 56). All amino acid positions are given relative to the reference sequence.
To detect frameshifts, DNA sequences from all patients were compared to the nef reference sequence. All nefs which had deletions due to frameshifts or premature stop codons (pseudo genes) were considered to be defective sequences and excluded from the main analysis. Only nondefective nefs, i.e., 153 P sequences from 22 patients, 615 LTNP sequences from 155 patients, and 2,311 UP sequences from 405 patients were considered for these main analyses (Table (Table2).2). The numbers of patients from different isolation sources are summarized in Table Table2.2. The primary comparison was done between LTNPs and Ps, and a secondary comparison, for reason of patient numbers, was made between LTNPs and UPs.
To find consensus sequence corresponding to a patient, all sequences of that patient were aligned by using MUSCLE (18, 19). The aligned sequences were used to generate consensus sequence using the “consensus” program from EMBOSS (46) that uses the sequence weights and a scoring matrix to calculate a score for each amino acid residue or nucleotide in the alignment; the highest-scoring residue goes into the consensus sequence if the score is higher than half the total weight of all of the sequences. For primary comparison, all consensus sequences from LTNPs and Ps were aligned by using the MUSCLE alignment program (18, 19). CLUSTAL X (54) was used to create phylogenetic trees using the neighbor-joining method (48). A total of 1,000 bootstrap replicates were performed in order to determine the confidence of the particular branching.
We then matched pairs of most closely related LTNP versus P patients, choosing the P patient whose consensus sequence was most closely related to the LTNP consensus from the phylogenetic tree of all LTNP and P consensus sequences. This was done as an attempt to reduce the amount of noise created by sequence changes during evolution that are not associated with the potential evolution of a switch in progression status. This matching of pairs was defined manually, based on inspection of the nef consensus sequence phylogenetic tree. A total of 20 pairs of LTNPs which were closely related to the Ps were obtained (Fig. (Fig.1).1). The details of the above 20 pairs of patients are provided in Table S2 in the supplemental material. For each pair, differences in residues were calculated and an observed score (SOi) was determined for each residue as defined by equation 1.
AL and NC represent alignment length and the number of pairs, respectively.
We wanted to determine whether the residue changes that defined the differences between a matched pair of LTNP and P consensuses were distributed differently compared to the differences seen with any pairwise combination of the 20 patients. Since we had matched the pairs to be closely related, then if there were subsequent amino acid changes occurring at a particular residue position altering the phenotype, such changes should be enriched among the matched pairs compared to the randomly matched pairs, which would then be expected to be in practically all cases more evolutionarily divergent. Thus, to calculate expected values of differences in residues, all combinations of 20 pairs of LTNP versus P pairs were generated, giving rise to 400 combinations. Subsequently, for each pair, similar to SOi, an expected score (SEi) was determined for each residue as defined by equation 3.
We then defined a score (Si) for each residue i as the ratio of observed and expected scores for that residue, as follows:
To visualize these findings along the nef sequence, we plotted histograms of SOi and SEi (Fig. (Fig.2).2). Scores for the most extreme residue (R1) and for the fifth most extreme residue (R5) were recorded as SR1 and SR5, respectively. In order to assess whether R1 andR5 were higher than expected by chance, we carried out permutation tests. A total of 10,000 permuted datasets were generated in which all 20 pairs of sequences were randomly assigned as being either Ps or LTNPs (i.e., typically ca. 50% of the pairs had their phenotypic status switched around), giving arise to 400 pairs. For all iterations, Si was recorded for each residue. For the 10,000 permutations, we then recorded in how many of them one or more residues with values greater than SR1 were observed and in how many permutations five or more residues with values greater than SR5 were observed.
The secondary comparison was done in between LTNP and UP and the whole process was repeated by replacing P by UP. We matched pairs of the most closely related LTNP versus UP patients, choosing the UP patient whose consensus sequence was most closely related to that of LTNP. A total of 47 closely related pairs were obtained. For each pair, differences in residues were calculated by using SOi. To calculate the SEi, all combinations of 47 pairs of LTNP versus P/UP pairs were generated, giving rise to 2,209 combinations. Score for the most extreme residue (SR1) and the fifth most extreme residue (SR5) were calculated, and a permutation test was carried out as explained above.
To study deletion patterns in nef sequences from all LTNP, P, and UP nucleotide sequences, consensus sequences corresponding to each patient were generated by using the consensus program EMBOSS (46). All consensus sequences, as reported in Table Table2,2, were aligned by using the MUSCLE alignment program (18, 19). The number of deletions for each codon was calculated. Then, for each codon, we performed the Fisher exact test to assess the level of association between the deleted/nondeleted genotype and the LTNP-versus-P or LTNP-versus-UP phenotype. All statistical analyses were performed by using the statistical package R (http://www.R-project.org). The Spearman's correlation statistics was used to find correlations between LTNP versus P or UP. Probabilities for comparison between LTNP and P or between LTNP and UP are denoted by PNP or PNU, respectively.
Since the Fisher exact test was calculated over every codon/residue of nef, we carried out a permutation test to determine the significance of the P value obtained from each residue's test. Similar to the replacement analysis, all consensus sequences were randomly assigned a P/UP or LTNP status, and the P values for each residue based on the Fisher exact test were calculated for each of the 1,000 simulations. The whole deletion study was repeated for the longest sequences corresponding to each patient. To find the longest sequence for a patient, we calculated the length of all of the sequences for that patient. The sequence with the maximum length out of the total sequences for a given patient was selected as the longest sequence.
LTNPs have a higher number of sequences containing frame shifts. The frameshifts appear to be distributed generally across both LTNPs and UPs (Fig. (Fig.3A).3A). There are two peaks of frameshifting in LTNPs occurring at 9SVIG (24 sequences) and 118QGYF (15 sequences) and one peak in UP patients at 8SSVI (10 sequences). In Ps, only one sequence had frameshifting at N52. We repeated the same for consensus (Fig. (Fig.3B)3B) and longest sequences (Fig. (Fig.3C)3C) per patient. In the case of LTNPs, there was one peak (S9) for both consensus sequences and longest sequences (two patients). Similarly, in UPs, there was one peak at 9SVI in both types of sequences, whereas in P patients no frameshifting was observed in case of consensus or longest sequences. Although it is possible that such frameshifting increases reflect selection maintaining an N-terminal activity while deleting a C-terminal one (e.g., see reference 9), it is difficult to assess significance, given that assessment of positional biases of frameshifts was not a predefined primary endpoint of our study. It is worth noticing that the length variations near the N terminus, as well as in some other parts of nef, are normal and usually do not have disruptive effects on the function of nef (11, 52).
Most of the LTNP patients sequenced were from the United States, Italy, and Australia, whereas the P patients were from the United States, Italy, and Japan (see Fig. S1 in the supplemental material). When we compare the total number of sequences obtained, there are many more defective nef genes among the LTNPs. LTNPs have a total of 149 defective nef sequences (20%) out of a total of 764 sequences, whereas Ps have 17 (10%) out of 170 and UPs have 315 (12%) out of a total 2,626 sequences. Ps and UPs combined have 332 defective nef genes (12%) out of a total 2,796 sequences, suggesting that defective nef genes are common among LTNPs. However, this may simply be an artifact of sampling, with a few LTNPs overinvestigated that have many defective nefs. The average lengths of nondefective nef genes in LTNPs, Ps, and UPs were 206.8, 207.1, and 207.7, respectively. A box plot of the lengths in all three categories is shown in Fig. Fig.4A.4A. We also studied percent defective proteins per patient in all three categories (see Table S1 in the supplemental material). The average number of patients sequenced in Ps is higher than for LTNPs and UPs (Table (Table2).2). However, when we calculated the average number of percent defective nef genes, we found that LTNPs had higher average number of percent defective nef genes (15.25%) than Ps (8.09%) and UPs (13.92%). The distribution of the percent defective nef genes in LTNPs and others is shown in Fig. Fig.4B4B.
However, the results presented above may be biased by the number of sequences investigated. When the analysis is performed on a patient-by-patient basis, there is no evidence that LTNPs have more defective nef genes (Table (Table3).3). Thus, there is no association between LTNP status and whether the patient has one or more nondefective nef genes (PNP = 0.23 and PNU = 0.39 [Fisher exact test]; Table Table3,3, footnote a). Similarly, there was no association between LTNP status and whether a patient has one or more defective nef genes (PNP = 0.47 and PNU = 0.052; Table Table3,3, footnote b). Thus, we conclude that there is no significant association between defective nef genes and disease progression in this data set.
Phylogenetic trees of all nondefective sequences from LTNPs and Ps or from LTNPs and UPs were generated. Each patient has sequences clustered together and clearly differentiable from those of other patients as described earlier (10). Consistent with previous studies (28, 29, 40, 43), visual inspection of this tree does not provide any strong evidence of phylogenetic grouping of LTNPs within the tree. We investigated more closely a set of LTNPs paired with closely related sequences from either Ps (Fig. (Fig.1)1) or UPs (see Fig. S2 in the supplemental material) to investigate amino acid replacements. The rationale behind this pairing was to try and focus on evolutionary changes that are associated with differences related to progression. This matching of pairs is intended to reduce some of the noise caused by other evolutionary changes that are not relevant to progression status, in the same way that a matched case-control study reduces the noise due to other covariates.
We examined the frequency of changes between the consensus sequence of LTNPs and the P consensus that they are paired with (as their nearest phylogenetically related sequence). We were interested in residues that show more differences between the consensus sequence of LTNPs and Ps. We found that certain residues show a much higher frequency of changes (Fig. (Fig.2).2). In particular, 13 residues—S8, V10, I11, A15, V85, V133, N157, S163, V168, D174, R178, E182, and R188—had high scores (SO > 0.015). The scores are normalized so that the evidence from all 20 pairs is treated equally, regardless of their phylogenetic distance. However, the scores obtained from inspection of the permuted pairs are also elevated for these residues (Fig. (Fig.2),2), implying that these residues are simply rapidly evolving residues. The correlation between SO and SE was significantly very high for all residues (ρ = 0.96; P < 2.2 × 10−16).
Inspection of Fig. Fig.22 reveals that one residue, N157, shows a relatively marked excess of observed change over expectation. To our knowledge, this residue has not been identified in previous studies as contributing to nonprogression. Other residues with a more modest excess included A15, V85, S163, V168, D174, R178, and R188. To test whether the excess of SO over SE seen for N157 or any other residues was significant, we did a more rigorous permutation test by looking at the number of permutations in which we observed a score as high as SR1 or SR5. Although some of these residues did show an excess of observed score over the expectation, the test revealed that a score at least as high as the highest score (SR1 = 2.28) occurred in 76.7% of the permuted datasets and that a score at least as high as the fifth highest score (SR5 = 1.48) occurred in 99.9% of the permuted datasets (for ranked residues, see Table S3 in the supplemental material). Thus, there is no evidence that any individual residue shows a significant excess of amino acid replacement between LTNP patients and their nearest related P patients.
We also investigated amino acid replacements between the consensus sequence of LTNP and the closest related UP consensus sequence (see Fig. S3 in the supplemental material). In all 47 pairs, we found that nine residues—S8, V10, I11, D28, K39, V133, S163, R178, and E182—had high scores (SO > 0.015). However, the scores obtained from the permuted pairs were also elevated, as observed in the case of previous 20 LTNP and P pairs. Similarly, the correlation between SO and SE was significantly very high for all residues (ρ = 0.98; P < 2.2 × 10−16). A more rigorous permutation test was done by looking at the number of permutations in which we observed a score as high as SR1 = 5.53 or SR5 = 1.84, and this revealed that scores as high as SR1 or SR5 were observed in all permutations.
We tested whether there was an excess of codon deletion overall among LTNPs. Table Table44 shows the number of patients with one or more deletions in all three categories of sequences. There is a slight but not significant (P = 0.83) excess of deletions among the LTNPs (26.45%) compared to UPs (26%) in case of consensus sequences. However, when only LTNPs and Ps are compared, the LTNPs seem to have fewer proteins with codon deletions compared to Ps (36%), but again these differences are not significant (P = 0.32).
When the analysis was repeated, replacing the consensus sequence for a patient with the longest sequence per patient (Table (Table4B),4B), there was still a slight excess of deletions among the LTNPs (27%) compared to UPs (26%), but this was not significant (P = 0.75). However, there was a lack of deletions among LTNPs compared to Ps (41%) only, but the Fisher exact test revealed that this finding was not significant (P = 0.21). Thus, LTNPs do not appear to have a markedly increased or decreased number of codon deletions overall.
We looked for deletions of specific regions of the Nef protein in all consensus sequences from LTNPs and others. Each codon present in the Nef NL4.3 sequence was compared to the codon present at that position in the alignment of all consensus sequences of LTNPs and other patients (either Ps or Ps and UPs). We calculated the probability (using the Fisher exact test) to determine whether there was a significant association between LTNP status and codon deletion for each amino acid along the nef sequence. The codons that had in-frame deletions in LTNPs are listed in Table Table5.5. Comparison of LTNPs and Ps revealed that four codons AGT (S9), GTG (V10), ATT (I11), and GCA (A49) were deleted in both Ps and LTNPs; however, only one codon GTG (V10) was found to be significantly deleted in LTNPs (P = 0.04). We also looked for deletions of specific regions of the Nef protein in all consensus sequences from LTNP and UP. The codons that had significant in-frame deletions are marked in boldface in Table Table5.5. These are located in the 152KVEEA region.
Since we performed multiple tests across each codon of nef, we considered whether such low P values are likely to arise by chance by carrying out permutation tests (see Materials and Methods). The appropriate cutoffs for significance (equivalent to P < 0.05) were 8.5 × 10−4 (LTNPs and Ps) and 8.1 × 10−5 (LTNPs and UPs), values lower than were seen for any of the observed residues (Table (Table5).5). We repeated the analysis described above by taking the longest sequence per patient instead of the consensus. The codons whose deletion is most significantly associated with LTNP (P < 1) are listed in Table S4 in the supplemental material. However, none of these codons were found to have P values lower than for those which occur by chance.
We found more defective/disrupted nef genes among LTNPs, as suggested by earlier studies (4, 10, 16, 33), apparently favoring the hypothesis that LTNPs have higher proportion of disruptive nef genes than Ps or UPs. However, this association was only seen when we looked at all sequences. When the analysis was carried out on a per-patient basis, the excess was not seen. The initial observation may reflect biased sampling in numbers of sequences per patient. Alternatively, later ascertainment of sequences from LTNPs postinfection may impact in some way on the intrapatient variability of the sequences observed. Therefore, we consider the evidence that there is an excess of defective sequences among LTNPs to be inconclusive. A prospective study would be required to address this question more fully. At the very least, this would require careful adjustment and control of the length of time postinfection among the comparison groups. Ideally, it would involve long-term follow-up with sequences from different disease stages to help avoid other potential confounding effects of the disease process on the mutation rate. Clearer and more consistent criteria for defining LTNP would also be beneficial, since the study presented here relied on definitions of nonprogression that are not identical across all studies. The study we present here can also be confounded by technical artifacts of the various study designs investigated, including bias in the number of sequences observed per patient, differing degrees of sequence variability per patient (which may partly relate to number of years postinfection), and variability of the sequencing techniques (Table (Table2).2). Such biases are easier to investigate in a positive association (to determine whether the positive association arises through such confounding of causative factors) than in a negative study such as ours. It is clear that such confounding factors in our study serve to somewhat reduce the power of the study to detect a true effect.
Other studies have suggested that amino acid replacements at particular residues are increased significantly among LTNPs (8, 10). However, in reviewing the evidence among the largest group of patients analyzed to date, we failed to confirm any of these findings or to detect any residues whose replacement is increased among LTNPs. Analysis of codon deletion revealed that a few regions of nef were more commonly deleted among LTNPs, in particular residues 152 to 156 (KVEEA) among nine patients (see Fig. S4 and Table S5 in the supplemental material). The region 152KVEEA is involved in β-COP recruitment (3, 21) and also predicted to be highly disordered (see Fig. S5 in the supplemental material), a feature shared by many regions that encode short signaling motifs (23). However, after a permutation test, we determined that this deleted region was not significant.
Although our study was focused on residue deletion and replacement, we did note an excess of frameshifting in the LTNPs occurring at specific regions of nef, which was not seen in the non-LTNP sequences (9SVIG and 118QGYF). This was not a primary endpoint in our study, and it is unclear exactly what statistical methods are most appropriate to address whether the excess in these regions represents a statistical departure from expectation. We can hypothesize that such frameshifting may be selected for in generating nef sequences that maintain the N-terminal activities while removing the C-terminal activities of the protein. However, analysis of an independent set of sequences would be required in order to formally test whether there is enrichment of frameshifting in these regions.
Thus, we can conclude that a significant proportion of nonprogression is unlikely to be attributable to amino acid replacement or deletion in specific regions of HIV-1 nef. Although we could not find any direct correlation between LTNP and P or UP based on the sequences studied, it is possible that, in addition to the fact that there are more deletions in nef from LTNPs, other parameters, such as the immune system response and virus fitness (other HIV gene products), could in combination with nef variability explain the outcome of other studies. Finally, it is also possible that rarer variants may contribute to the disease progression in individual patients.
This study was funded by Science Foundation Ireland and University College Dublin, Dublin, Ireland.
We thank Cathal Seoighe (NUI, Galway, Ireland) for helpful comments. We also thank two anonymous reviewers for their comments and suggestions, which helped us in improving the manuscript.
Published ahead of print on 13 January 2010.
†Supplemental material for this article may be found at http://jvi.asm.org/.