Download Positively selected sites on the surface glycoprotein (G) of infectious

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Expression vector wikipedia , lookup

Interactome wikipedia , lookup

Protein wikipedia , lookup

Metalloprotein wikipedia , lookup

Amino acid synthesis wikipedia , lookup

Plant virus wikipedia , lookup

Vectors in gene therapy wikipedia , lookup

Biosynthesis wikipedia , lookup

Western blot wikipedia , lookup

Protein–protein interaction wikipedia , lookup

Proteolysis wikipedia , lookup

Homology modeling wikipedia , lookup

Ancestral sequence reconstruction wikipedia , lookup

Biochemistry wikipedia , lookup

Two-hybrid screening wikipedia , lookup

Genetic code wikipedia , lookup

Point mutation wikipedia , lookup

Transcript
Journal of General Virology (2008), 89, 703–708
DOI 10.1099/vir.0.83451-0
Positively selected sites on the surface glycoprotein
(G) of infectious hematopoietic necrosis virus
Scott E. LaPatra,1 Caryn Evilia2 and Vern Winston2
Correspondence
1
Vern Winston
2
Clear Springs Foods Inc., PO Box 712, Buhl, ID 83316, USA
Department of Biological Sciences, Campus Box 8007, Idaho State University, Pocatello,
ID 83209-8007, USA
[email protected]
Received 17 September 2007
Accepted 1 November 2007
Mutations in the surface glycoprotein (G) of infectious hematopoietic necrosis virus (IHNV), a
rhabdovirus that causes significant losses in hatcheries raising salmonid fish, were studied. A
303 nt segment (mid-G region) of this protein from 88 Idaho isolates of IHNV was sequenced.
Evidence of positive selection at individual codon sites was estimated by using a Bayesian
method (MrBayes). A software algorithm (CPHmodels) was used to construct a threedimensional (3D) representation of the IHNV protein. The software identified structural
homologies between the IHNV G protein and the surface glycoprotein of vesicular stomatitis virus
(VSV) and used the VSV structure as a template for predicting the IHNV structure. The amino
acids predicted to be under positive selection were mapped onto the proposed IHNV 3D
structure and appeared at sites on the surface of the protein where antigen–antibody interaction
should be possible. The sites identified as being under positive selection on the IHNV protein
corresponded to those reported by others as active sites of mutation for IHNV, and also as
antigenic sites on VSV. Knowledge of the sites where genetic variation is positively selected
enables a better understanding of the interaction of the virus with its host, and with the host
immune system. This information could be used to develop strategies for vaccine development for
IHNV, as well as for other viruses.
INTRODUCTION
Infectious hematopoietic necrosis virus (IHNV) is an
important rhabdoviral pathogen of salmonid fishes (Tordo
et al., 2005), causing a large economic impact on
commercial fish farms as well as hatcheries raising fish
for restocking and mitigation efforts (http://usda.mannlib.
cornell.edu/usda/current/TrouProd/TrouProd-02-26-2007.
pdf). Because of its economic importance, IHNV has been
the subject of intense study. Much of this effort has been
directed toward an understanding of the evolution of the
virus (Nichol et al., 1995; Oshima et al., 1995; Huang et al.,
1996; Emmenegger et al., 2000; Troyer et al., 2000; Kurath
et al., 2003; Troyer & Kurath, 2003), with particular focus
on identifying the sites at which the virus proteins change
as they evolve.
Because of physical and functional constraints, very few
mutations result in an increase in virus fitness (Domingo,
2006). Mutations that result in decreased fitness of the
virus are removed from the gene pool by negative selection.
Some (if not most) changes are neutral. They have no
negative effect on fitness, but also do not provide a selective
advantage (Domingo, 2006). In the case of virus surface
proteins, those that demonstrate enhanced fitness might
behave in a number of ways. In one instance, the change
might result in an improved interaction with the host, by
0008-3451 G 2008 SGM
more efficient host binding, entry or uncoating of the virus.
Alternatively, the changes might disrupt the interaction of
the virus with the proteins of the host immune system.
Specific antibodies or receptors on the cells of the immune
system might not recognize the altered epitopes as
effectively, resulting in an enhanced ability of the virus to
escape the defence systems of the host. These sites are
under positive selection and are identifiable because the
number of non-synonymous amino acid changes at these
sites exceeds the number of synonymous changes
(Domingo, 2006; Yang et al., 2000). Until the advent of
high-speed computers, identification of sites of positive
selection was not reliable. However, with the widespread
availability of high-speed computers, it has been possible to
develop methods to identify individual sites under positive
selection. Two of the most commonly used programs to
identify sites of positive selection are PAML (Yang, 2007)
and MrBayes (Huelsenbeck & Dyer, 2004). In our study,
recent (1990–2006) isolates of IHNV were obtained from
commercial fish farms in the state of Idaho, USA. A 303 nt
segment of the major surface (G) protein gene was
amplified and sequenced. The sequences were evaluated
by using a fully Bayesian method (MrBayes; Huelsenbeck &
Dyer, 2004) to identify codons where the rate of nonsynonymous mutation exceeded that of synonymous
mutation in a manner consistent with positive selection.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 11 May 2017 23:48:56
Printed in Great Britain
703
S. E. LaPatra, C. Evilia and V. Winston
To test the relevance of these predictions, attempts were
made to correlate the location of positively selected sites on
the amino acid sequence with the three-dimensional (3D)
structure of the virus protein. As the structure of the IHNV
G surface protein has not been determined, we used a
software algorithm (CHPmodels; Lund et al., 2002) to
construct a 3D representation of the IHNV G protein. This
approach utilized the structural and sequence homologies
between the IHNV G sequence and the vesicular stomatitis
virus (VSV) surface protein sequence (Roche et al., 2007)
to construct a model of the IHNV molecule. The locations
of positively selected sites were mapped onto the predicted
3D model of the IHNV protein. The predicted sites on the
IHNV sequence mapped onto the surface of the protein
that would be expected to be in contact with antibodies
and/or cellular receptors, and at sites reported by others
(Huang et al., 1996; Troyer et al., 2000) to be sites of
mutation in the IHNV molecule. These regions also
corresponded to the major antigenic sites of the VSV
surface protein (Vandepol et al., 1986).
METHODS
Source of virus. Isolates of IHNV were collected from outbreaks at
commercial fish farms located in Idaho, USA, over the course of
16 years (1990–2006). Samples were inoculated into cell culture (EPC
or CHSE-214 cells) and identified as IHNV by virus neutralization.
Viral lysates were stored at 275 uC. None of the isolates were
passaged more than three times in culture.
2.0 homology modelling server (http://www.cbs.dtu.dk/services/
CPHmodels/) (Lund et al., 2002). The amino acid sequence of the
IHNV G protein (GenBank accession no. AAC42146; SRCV strain)
was used as input to the web interface. The software identified the B
chain of the VSV surface protein (2CMZ.pdb) as the highest-scoring
template candidate, and constructed a 3D representation of the IHNV
protein based on this template. The locations of positively selected
sites were visualized by using VMD software (Visual Molecular
Dynamics). A CLUSTAL W alignment (Thompson et al., 1994) was
used to confirm the correspondence between IHNV and VSV
sequences (data not shown).
RESULTS AND DISCUSSION
Sequence analysis
In total, 88 sequences were obtained. Of these, seven were
unique. The rest were observed between two and 31 times
(Table 1). Representatives of each of these sequence groups
were aligned by codon and analysed by using MrBayes. To
validate the software used in these experiments, preliminary experiments used MrBayes to evaluate datasets of
influenza HA protein (haemagglutinin), human immunodeficiency virus envelope protein and b-globin for evidence
of positive selection. The results obtained were in strong
agreement with those obtained by other computational
approaches (Yang et al., 2000; Huelsenbeck & Dyer, 2004)
(data not shown). The consensus tree obtained by using
MrBayes on the IHNV sequence is shown in Fig. 1. This
tree was in agreement with the trees describing the
Isolation of RNA, RT-PCR and sequencing. RNA was isolated from
cell lysates by using a QIAamp viral RNA mini kit (Qiagen) following
the manufacturer’s instructions. DNA was synthesized from viral RNA
by using an RT-PCR kit (Qiagen One-Step RT-PCR) as directed by the
manufacturer, using outer primers described by Emmenegger et al.
(2000). Before sequencing, PCR primers were removed by using
ExoSAP shrimp alkaline phosphatase (USB). PCR products were
sequenced by using an ABI 3100 Genetic Analyzer (Idaho State
University Molecular Research Core Facility), BigDye chemistry (ABI)
and the inner primers described by Emmenegger et al. (2000) as
sequencing primers. Each PCR product was sequenced in both
directions and the output was analysed by using the Staden Package
(Staden et al., 2000) for evaluation of base calls and production of a
contiguous alignment of the complementary sequence.
Sequence analysis. Sequences were aligned by codon and trimmed
to the 303 nt mid-G sequence reported by others (Emmenegger et al.,
2000; Troyer et al., 2000; Kurath et al., 2003; Troyer & Kurath, 2003).
Reference sequences included GenBank accession numbers AF237983–
AF237992, which represent earlier Idaho isolates (Troyer et al., 2000),
L40878, a representative of M clade isolates (Nichol et al., 1995), and
L40881 (SRCV), used by others (Troyer & Kurath, 2003) as an
outgroup sequence. Bayesian analysis was performed by using MrBayes
3.1.2 (Huelsenbeck & Dyer, 2004) in parallel mode. Specific commands
for MrBayes were: ‘lset nucmodel5codon nst52 omegavar5m3’,
‘report possel5yes’. Each chain was run for 1.26106 cycles. The sump
and sumt commands were used to tabulate posterior probabilities of
positive selection of each amino acid site, and to build consensus trees.
Results of estimations obtained before the process reached convergence
were discarded. Typically, the first 200 000 cycles were discarded.
Mapping of IHNV sites on VSV 3D structure. The 3D structure of
Table 1. Representatives of each of the sequence groups and
frequencies of occurrence among the 88 sequences obtained
The Fa18 sequence is identical to that of GenBank accession no.
AF237991. The Fw38 sequence is identical to that of GenBank
accession no. AF237983. The Fw40 sequence is identical to that of
GenBank accession no. AF237987 (WRAC strain).
Representative
isolate
Fa1
Fa2
Fa6
Fa9
Fa11
Fa13
Fa18
Fw4
Fw6
Fw7
Fw12
Fw14
Fw33
Fw34
Fw35
Fw38
Fw40
GenBank
accession no.
EU249526
EU249527
EU249528
EU249529
EU249524
EU249525
EU249539
EU249535
EU249536
EU249537
EU249530
EU249531
EU249532
EU249533
EU249534
EU249538
EU249540
No. occurrences
1
9
1
22
2
2
1
1
4
1
2
31
1
1
4
2
3
the IHNV G protein was predicted by using the CPHmodels
704
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 11 May 2017 23:48:56
Journal of General Virology 89
Positively selected sites on G protein of IHNV
Fig. 1. Tree produced by MrBayes using nonredundant Idaho sequences (indicated by a
prefix of Fw or Fa). GenBank sequences with
accession numbers AF237983–AF237992,
L40881 (outgroup) and L40878 are also
included. Suffixes following sequence names
indicate the number of times that the sequence
was observed. Clade credibility values (%) are
shown above branches.
relationships of Idaho strains reported by others (Troyer
et al., 2000; Troyer & Kurath, 2003) in the following respects:
(i) all (except the outgroup) were descended from group M
strains; (ii) the two major clades that were observed were
divided so that clade A–B contained the reference sequences
associated with subgroups A and B of group M as previously
reported (Troyer et al., 2000); and (iii) clade C–D contained
the reference sequences associated with subgroups C and D
reported by the same authors (Troyer et al., 2000).
Estimates of mean posterior probability that a codon was
under positive selection are represented in Table 2, Fig. 2
and Fig. 3(a–c). When all of the sequences (clade A–B and
clade C–D) represented in Fig. 1 were analysed together,
the amino acids with a mean probability of being under
positive selection .95 % were aa 252, 256, 270, 272 and
277 (using the numbering scheme for the whole protein)
(Table 2; Figs 2a, 3a). Inspection of the results showed a
Table 2. Correlation between amino acid sites and posterior
probabilities of positive selection
–, Values ,70 %.
Amino
acid
220
247
251
252
256
270
272
277
284
Probability of positive selection
Clade A–B and C–D
0.746178
0.812231
0.776420
0.964077
0.999696
0.990682
0.968439
0.979865
0.728598
http://vir.sgmjournals.org
Clade A–B
Clade C–D
0.955825
–
–
–
0.993278
0.994983
0.989305
0.964256
0.964192
–
0.956028
0.884004
0.997489
0.998601
0.915735
0.960275
0.878237
–
number of sites with mean probabilities .70 % of being
under positive selection. These sites were 220, 247, 251 and
284 (Table 2; Figs 2a, 3a). These results are in general
agreement with those of others. Huang et al. (1996)
reported that changes at aa 78, 81, 230–231, 272–273 and
275–276 (Fig. 3e) of IHNV enabled mutants to escape
neutralizing monoclonal antibodies. In another study,
Troyer et al. (2000) reported observing non-synonymous
mutations at IHNV aa 252, 256, 270, 275–277, 284 and 285
(Fig. 3f). The major difference between our study and that
of Huang et al. (1996) was their observation of mutations
at positions 230 and 231. We saw no amino acid changes at
these positions in any of our sequences. Because we did not
sequence the whole molecule, we could not observe the
state of positions 78 and 81. It has been suggested that
passage in cell culture may select for anomalous changes in
virus proteins (Novella et al., 2005). The fact that our
isolates had not been passaged extensively in culture may
explain why we did not observe changes at these positions.
The closer agreement between our results and those of
Troyer et al. (2000) may be because both that study and
ours used isolates that had not been passaged extensively.
As it has been suggested previously (Huelsenbeck et al.,
2006) that patterns of selection can be different in different
lineages, the sequences in each of the main clades shown in
Fig. 1 were analysed separately by using MrBayes. The
results obtained by using the sequences in clade A–B are
shown in Table 2, Fig. 2(b) and Fig. 3(b). In this case, the
amino acids with mean posterior probabilities of positive
selection .95 % were 220, 256, 270, 272, 277 and 284. No
other sequences with a mean posterior probability .70 %
were observed. The results when the sequences of clade
C–D were analysed separately are shown in Table 2,
Fig. 2(c) and Fig. 3(c). In this case, the amino acids with
mean posterior probabilities of positive selection .95 %
were 247, 252, 256 and 272. Amino acids with mean
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 11 May 2017 23:48:56
705
S. E. LaPatra, C. Evilia and V. Winston
cells, the sites involved must be located on the outside
surface of the virus so that these interactions can occur. To
visualize the possible location of positively selected sites on
the virus, the 3D modelling software CPHmodels (Lund
et al., 2002) was used. This software identified the B chain
of the surface protein of VSV (2CMZ.pdb) as the highestscoring template candidate. A 3D model of the IHNV
protein was constructed by using the VSV protein as a
template. The resulting structures are shown in Fig. 3(a–c),
with the corresponding prefusion form of the VSV
structure shown in Fig. 3(d) [note that, in Fig. 3(a–c),
only the mid-G portion of the IHNV molecule is
represented]. In this figure, the molecules are oriented so
that the view is from the top, facing directly toward the
membrane. Fig. 3(a–c) are labelled to represent sites
identified as being under positive selection at the .95 %
(red) and .70 % (green) confidence levels by MrBayes.
Fig. 3(a) represents sites identified when all members of the
dataset were included. Fig. 3(b, c) represent sites identified
when only the members of clade A–B or clade C–D,
respectively, were used in the analysis. Fig. 3(d) is labelled
to reflect the locations of amino acids that change in
monoclonal antibody escape mutants of VSV (Vandepol
et al., 1986). Fig. 3(e, f) represent the result of mapping
sites of IHNV G amino acid changes reported by others
(Huang et al., 1996; Troyer et al., 2000) on the prefusion
form of the molecule.
Fig. 2. Histograms of posterior probabilities that amino acids are
under positive selection, obtained by using (a) all of the sequence
data that produced the tree shown in Fig. 1; (b) the sequences in
clade A–B (Fig. 1); (c) the sequences in clade C–D (Fig. 1). See
Table 2 for values.
posterior probabilities .70 % were 251, 270 and 277
(Table 2; Figs 2c, 3c.). Further research is needed to
determine whether these changes in the virus protein
provide the mutated virus with a selective advantage.
Structural comparisons
If positive selection is arising as a result of the interaction
between virus and host, either as a function of antigen–
antibody interaction or as a result of enhanced binding to
706
All of the IHNV G sites identified by the Bayesian approach
as undergoing positive selection were on or near the top
surface of the molecule. This is consistent with the
hypothesis that these sites may be involved in interaction
with host antibodies. The cluster of IHNV G sites from
aa 270 to 277 is in a prominent a-helix (helix E), which
should be readily accessible to antibodies. This region
corresponds to the VSV A2 region, which is one of the two
major epitopic regions of that virus (Fig. 3d). Helix E is
also in a region of the molecule that does not change shape
as the molecule converts from its prefusion state in the
extracellular virion to the pH-activated state in the
lysosome (Roche et al., 2007). This would allow for more
amino acid substitutions in this region. Amino acid
changes in a hinge region, for example, would be more
damaging to the function of the protein. Notable is aa 274,
which is conserved in all of these sequences. This could
indicate that this site is critical for the stabilization of this
helix or for the binding of the virus to the host cell. The
pattern of sites that were identified as undergoing positive
selection by the Bayesian approach was in general
agreement with reports of others (Huang et al., 1996;
Troyer et al., 2000) (Fig. 3e, f). However, Huang et al.
(1996) observed variation at aa 230 and 231 but, in our
study, these sites were absolutely conserved. The difference
between our study and theirs could be explained by the fact
that, in their study, the virus had been passaged repeatedly
in culture to produce antibody-escape mutants. It is known
that repeated passage in culture can allow amino acid
changes that are not observed in vivo (Novella et al., 2005).
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 11 May 2017 23:48:56
Journal of General Virology 89
Positively selected sites on G protein of IHNV
Fig. 3. Visualization of the possible locations of positively selected sites, obtained by using the 3D modelling software
CPHmodels (Lund et al., 2002). (a) Clades A–B and C–D. (b) Clade A–B analysed alone. (c) Clade C–D analysed alone. For
(a–c), the IHNV amino acids with a probability of .95 or .70 % (MrBayes) of undergoing positive selection are represented by
red and green areas, respectively. (d) VSV epitopes (VSV numbering). (e) IHNV amino acids identified by Huang et al. (1996) as
providing resistance to antibody neutralization. (f) IHNV sites identified by Troyer et al. (2000) as sites of non-synonymous
mutation.
Troyer et al. (2000) reported no changes at these sites,
which may reflect the fact that the virus used in that study
had not been passaged extensively. Finally, the A1 epitopic
region of the VSV structure appeared to share the same
general region of the molecule as aa 78 and 81 of IHNV G,
identified by Huang et al. (1996) as an epitope of IHNV.
http://vir.sgmjournals.org
This would suggest that it might be informative if future
studies also sequenced this region of this molecule.
Further work is needed to explore the suggestions provided
by these results. A determination of the 3D structure of the
IHNV G protein is needed to confirm the location of the
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 11 May 2017 23:48:56
707
S. E. LaPatra, C. Evilia and V. Winston
positively selected sites on the surface of the protein
identified by the Bayesian approach. If it is possible to
identify the regions of the G protein that are undergoing
rapid selection, it might be possible to design vaccines
whose sequences mirror the specific patterns of change
being observed. Conversely, the fact that areas of the
protein are conserved may imply that change in these areas
is impossible if the virus is to remain viable. Vaccines
directed toward these vital regions might be more effective,
because the virus is prevented by structural constraints
from mutating in these areas.
Lund, O., Nielsen, M., Lundegaard, C. & Worning, P. (2002).
CPHmodels 2.0: X3M a computer program to extract 3D models.
In Abstracts of the CASP5 conference (the Fifth Community Wide
Experiment on the Critical Assessment of Techniques for Protein
Structure Prediction), Asilomar, CA, USA, 2002, A102. http://
www.cbs.dtu.dk/services/CPHmodels/abstract.php
Nichol, S. T., Rowe, J. E. & Winton, J. R. (1995). Molecular
epizootiology and evolution of the glycoprotein and non-virion
protein genes of infectious hematopoietic necrosis virus, a fish
rhabdovirus. Virus Res 38, 159–173.
Novella, I. S., Gilbertson, D. L., Borrego, B., Domingo, E. & Holland,
J. J. (2005). Adaptability costs in immune escape variants of vesicular
stomatitis virus. Virus Res 107, 27–34.
Oshima, K. H., Arakawa, C. K., Higman, K. H., Landolt, M. L., Nichol,
S. T. & Winton, J. R. (1995). The genetic diversity and epizootiology of
ACKNOWLEDGEMENTS
This work was partially supported by NIH grant P20 RR16454 from the
Biomedical Research Infrastructure Network/Idea Network of
Biomedical Research Excellence BRIN/INBRE Program of the
National Center for Research Resources. The authors acknowledge
Luobin Yang for his assistance with software, and Gael Kurath and Ryan
Troyer for sharing unpublished sequences. Eric Anderson provided
careful reading and helpful comments on the manuscript. George
Vidaver provided important guidance on protein structure questions.
REFERENCES
Domingo, E. (2006). Virus evolution. In Fields Virology, 5th edn,
pp. 389–406. Edited by D. M. Knipe & P. M. Howley. Baltimore, MD:
Lippincott Williams & Wilkins.
Emmenegger, E. J., Meyers, T. R., Burton, T. O. & Kurath, G. (2000).
Genetic diversity and epidemiology of infectious hematopoietic
necrosis virus in Alaska. Dis Aquat Organ 40, 163–176.
Huang, C., Chien, M. S., Landolt, M., Batts, W. & Winton, J. (1996).
Mapping the neutralizing epitopes on the glycoprotein of infectious
haematopoietic necrosis virus, a fish rhabdovirus. J Gen Virol 77,
3033–3040.
Huelsenbeck, J. P. & Dyer, K. A. (2004). Bayesian estimation of
positively selected sites. J Mol Evol 58, 661–672.
Huelsenbeck, J. P., Jain, S., Frost, S. W. D. & Kosakovsky-Pond, S. L.
(2006). A Dirichlet process model for detecting positive selection
infectious hematopoietic necrosis virus. Virus Res 35, 123–141.
Roche, S., Rey, F., Gaudin, Y. & Bressanelli, S. (2007). Structure of
the prefusion form of the vesicular stomatitis virus glycoprotein G.
Science 315, 843–848.
Staden, R., Beal, K. F. & Bonfield, J. K. (2000). The Staden package,
1998. Methods Mol Biol 132, 115–130.
CLUSTAL W:
improving the sensitivity of progressive multiple sequence alignment
through sequence weighting, position-specific gap penalties and
weight matrix choice. Nucleic Acids Res 22, 4673–4680.
Thompson, J. D., Higgins, D. G. & Gibson, T. J. (1994).
Tordo, N., Benmansour, A., Calisher, C., Dietzgen, R. G., Fang, R.-X.,
Jackson, A. O., Kurath, G., Nadin-Davis, S., Tesh, R. B. & Walker, P. J.
(2005). Family Rhabdoviridae. In Virus Taxonomy: Eighth Report of the
International Committee on Taxonomy of Viruses, pp. 623–644. Edited
by C. Fauquet, M. A. Mayo, J. Maniloff, U. Desselberger & L. A. Ball.
San Diego, CA: Academic Press.
Troyer, R. M. & Kurath, G. (2003). Molecular epidemiology of
infectious hematopoietic necrosis virus reveals complex virus traffic
and evolution within Southern Idaho aquaculture. Dis Aquat Organ
55, 175–185.
Troyer, R. M., Lapatra, S. E. & Kurath, G. (2000). Genetic analyses
reveal unusually high diversity of infectious haematopoietic necrosis
virus in rainbow trout aquaculture. J Gen Virol 81, 2823–2832.
Vandepol, S. B., Lefrancois, L. & Holland, J. J. (1986). Sequences of
the major antibody binding epitopes of the Indiana serotype of
vesicular stomatitis virus. Virology 148, 312–325.
in protein-coding DNA sequences. Proc Natl Acad Sci U S A 103,
6263–6268.
Yang, Z. (2007).
Kurath, G., Garver, K. A., Troyer, R. M., Emmenegger, E. J., EinerJensen, K. & Anderson, E. D. (2003). Phylogeography of infectious
Yang, Z., Nielsen, R., Goldman, N. & Pedersen, A.-M. (2000). Codon-
haematopoietic necrosis virus in North America. J Gen Virol 84,
803–814.
substitution models for heterogeneous selection pressure at amino
acid sites. Genetics 155, 431–449.
708
PAML 4: phylogenetic analysis by maximum
likelihood. Mol Biol Evol 24, 1586–1591.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 11 May 2017 23:48:56
Journal of General Virology 89