Hindawi Publishing Corporation

Computational and Mathematical Methods in Medicine

Volume 2015, Article ID 395826, 9 pages
http://dx.doi.org/10.1155/2015/395826

Review Article
Genetic Consequences of Antiviral Therapy on HIV-1
Miguel Arenas1,2
1

Institute of Molecular Pathology and Immunology of the University of Porto (IPATIMUP), 4200-465 Porto, Portugal
Centre for Molecular Biology Severo Ochoa, Consejo Superior de Investigaciones Cientficas (CSIC), 28049 Madrid, Spain

Correspondence should be addressed to Miguel Arenas; [email protected]

Received 22 February 2015; Revised 26 May 2015; Accepted 27 May 2015
Academic Editor: Qi Dai
Copyright 2015 Miguel Arenas. This is an open access article distributed under the Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
A variety of enzyme inhibitors have been developed in combating HIV-1, however the fast evolutionary rate of this virus commonly
leads to the emergence of resistance mutations that finally allows the mutant virus to survive. This review explores the main
genetic consequences of HIV-1 molecular evolution during antiviral therapies, including the viral genetic diversity and molecular
adaptation. The role of recombination in the generation of drug resistance is also analyzed. Besides the investigation and discussion
of published works, an evolutionary analysis of protease-coding genes collected from patients before and after treatment with
different protease inhibitors was included to validate previous studies. Finally, the review discusses the importance of considering
genetic consequences of antiviral therapies in models of HIV-1 evolution that could improve current genotypic resistance testing
and treatments design.

1. Introduction
According to UNAIDS, the Joint United Nations Programme
on HIV/AIDS World Health Organization, a total of 35.3
[32.238.8] million people worldwide were living with HIV-1
in 2012, indicating a 15% increase of infected people from
2001 [1]. A total of 2.3 [1.92.7] million were newly infected
during 2012, showing a 33% decline of new infections from
2001 with 3.4 [3.13.7] million. Indeed, the number of AIDS
deaths declined from 2.3 [2.12.6] million in 2005 to 1.6
[1.41.9] million in 2012 [1]. An important cause for such a
death decline is the antiretroviral therapy, usually referred to
as highly active antiretroviral therapy (HAART). In 2012, a
total of 9.7 million people from low/middle-income countries
received HAART and the UNAIDS expects to reach 15 million
people receiving HAART by 2015 [1]. Nevertheless, in 2013
only 34% of people infected with HIV in low/middle-income
countries (28.6 million) could receive therapy [1]. Therefore,
there are still important regional differences that should be
solved [2, 3]. On the other hand, the development of an
effective HIV vaccine is still under progress with a number
of failures [4] because of the high rate of evolution of HIV1 [5, 6]. As a consequence, up to date the only treatment for
HIV-1 is the antiretroviral drug therapy.

HAART have largely delayed the onset of AIDS-related

illness and death [1] although they cannot eradicate the virus
mainly due to latent viral reservoirs [7]. In addition, drug
resistance mutations can reduce the activity of the therapy
[8, 9]. Drug resistance mutations probably emerge because
HIV evolves rapidly, with high mutation and recombination
rates and under rapid population dynamics [10]. Of course,
then natural and drug-induced selection can eliminate
most of viral variants [11]. The surviving variants (820%)
present drug resistance mutations, which allows recovering
fitness and replication capacity [8, 12]. Interestingly, different
inhibitors can generate different selective pressures that
induce the fixation of different resistance mutations in the
viral population but also different resistance mutations may
affect different inhibitors in a different fashion. This suggests
the simultaneous use of more than one inhibitor that could
cover a wider range of mutations [13], although this strategy
may fall into similar resistance (cross-resistance) and lack of
synergy [14, 15].
A potential strategy to deal with the problem of resistance mutations could be the consideration of the molecular
evolution of the virus [16, 17] into the inhibitor design. For
example, inhibitors that account for molecular evolutionary
processes of the virus could eliminate viral variants that could

2
be predicted beforehand. Actually, this promising strategy is
commonly applied to HIV-1 vaccines design through the use
of centralized (consensus, center-of-tree or ancestral) genes
that can induce immune responses (reviewed in [17]). Such
centralized sequences could consider the immunogenetic
particularities of the diverse circulating variants in the target
population [18, 19]. However, although these centralized
vaccines generated promising antibody responses, they were
only partially effective in covering the large HIV-1 genetic
diversity. Perhaps this could be derived from the application of not enough realistic models of HIV-1 evolution
as suggested in [17]; see also [20]. Knowledge on HIV-1
molecular evolution can also be used to develop realistic
models of evolution [21, 22] that can be applied for additional
purposes such as the prediction of resistance mutations [23]
or the evolutionary reply of the viral population, genotypic
resistance testing [24, 25].
This study explores the genetic consequences of antiviral
therapy on HIV-1. First, it analyzes the influences of antiviral
therapies on the viral genetic diversity, including the particular roles of the substitution and recombination processes in
the generation of drug resistance. Then, the molecular signatures of selective pressures derived from antiviral therapies
are evaluated. A brief evolutionary analysis of the influence
of different protease (PR) inhibitors (PIs) on the PR-coding
region was performed to evaluate previous works and to
provide an illustrative example. The application of the genetic
consequences derived from antiviral therapies in the development of new empirical substitution models that could be
used for purposes such as genotypic resistance testing and
treatments design is also discussed.

2. Genetic Diversity Generated during

HIV-1 Antiviral Therapy
Interestingly, the effects of HIV-1 antiviral drugs on the viral
genetic diversity depend on the evolutionary level under
study. It differs from overall diversity of circulating strains in
the viral population to local nucleotide diversity of particular
viral genes.
The antiviral therapy can reduce the global viral genetic
diversity in the population due to the selection of viral strains
[2628]. This phenomenon can be interpreted as a classical population range contraction and habitat fragmentation
that commonly tend to decrease genetic diversity [29, 30].
Actually, a variety of population genetics analysis of HIV-1
showed the existence of severe population bottlenecks (loss
of viral load) and loss of virus fitness during drug regimens
[28, 31, 32].
In contrast, the survival strains may present drug resistance mutations [33] that often increase genetic diversity of
the protein-coding genes of the target proteins [12, 34, 35].
Wu et al. [36] found that patients treated with several PIs
presented 3 times more protease mutations than untreated
patients. These findings are also observed in the computational analysis presented in the last section of this paper
where most of PIs promoted higher levels of nucleotide diversity in the PR-coding gene. Interestingly, pairs and clusters

Computational and Mathematical Methods in Medicine

of correlated resistance mutations (coevolution) were significantly more abundant in treated patients [36]. Consequently,
the increased diversity does not follow a random process,
instead the new mutations present residue-residue interactions from direct association with viral protein inhibitors
[23]. Increased genetic diversity can also be observed under
treatment with other antiviral drugs such as reverse transcriptase (RT) inhibitors [34, 3739] and integrase (IN)
inhibitors [4043], although the increased diversity under
the latter drug class is mainly based on secondary resistance
mutations [4043]. Indeed, combinations of different drug
classes (acting on different HIV-1 proteins) can generate
synergistic inhibition [44] but the overall presence of synergy
on genetic diversity remains to be explored, although some
mutations in the Env region have already been associated with
resistance to entry inhibitors that affect other viral genes [9].
Overall, at this level, the molecular mechanisms by which
the virus can evade treatments seem directly related with
the viruss ability to generate genetic diversity in a particular
environment. Thus, this increased genetic diversity could be
driven by strong selective pressures (discussed later).

3. The Role of Viral Recombination during

HIV-1 Antiviral Therapy
Recombination constitutes a fundamental evolutionary force
in HIV generating new viral strains, increasing viral diversity,
and facilitating adaptation [4547]. Indeed, ignored recombination can bias the inference of a variety of evolutionary
processes and parameters (i.e., it can increase the number of
false positively selected sites [48, 49] or generate incorrect
phylogenetic tree and ancestral sequence reconstructions [50,
51]). Therefore recombination should be taken into account
for analyzing and understanding HIV-1 evolution.
The role of recombination on the emergence of drug resistance mutations is not yet clear and it can be difficult to assess
because other processes may also influence its evolutionary
consequences (i.e., cellular superinfection [5254], random
genetic drift, and viral population size [55, 56] or fitness
selection of the newly generated viral forms [57, 58]) and
because the detection of recombination can be problematic
under low levels of nucleotide diversity [59]. Contradictory
effects of recombination during HIV-1 antiviral therapy can
be found in the literature.
As one would expect beforehand, several studies showed
that recombination is crucial to generate drug resistance.
A computer simulations study suggested that recombination might favor the generation of drug resistance [60]. In
addition, HIV-1 strains derived from recombination events
presented resistance mutations [61, 62].
On the contrary, Archer et al. [63] showed that despite
the wide diversity of recombinant forms in HIV populations,
only a minority of recombination events are of significance to
the evolution of the virus. Counterintuitively, it has also been
demonstrated that recombination can slow down the generation of multi-drug-resistant strains during therapy [52] and
it may be suppressed by selection for resistance to PIs [64].
It seems that the initial genetic barrier caused by recombination (most of recombinant forms could present low

Computational and Mathematical Methods in Medicine

fitness) could reduce the fitness of the viral population during
the therapy but in case a recombinant form is selected,
resistance mutations could be better able to persist in the viral
population [54] and speed up adaptation (the Fisher-Muller
effect) [65]. In any case, these opposite findings suggest
that more sophisticated analyses should be performed to
determine the influence of recombination on the emergence
of drug resistance mutations, as suggested by Shi et al. [61].

4. Selective Pressures Induced by

HIV-1 Antiviral Therapy
Antiviral therapy may cause important selective pressures on
viral populations [12, 66]. In particular, severe fitness losses
can be derived from antiviral treatments until the emergence
of beneficial mutations that allow restoring the vital replication capacity [12]. Thus, resistance to viral inhibitors can drive
the fixation of favorable variants [23, 67].
The overall response to antiviral drugs presented an
excess of nonsynonymous substitutions [23, 68] (which was
also found in the analysis presented in the following section).
For example, Wu et al. [36] found that an antiviral therapy
can induce diversifying selection in nearly one-half of PR
sites. It is widely known that positively selected sites (PSSs)
are often located in the protein surface, whereas conserved
or negatively selected sites (NSSs) are commonly observed
in the protein core in order to conserve the protein function
[69]. However, the molecular adaptation induced by antiviral
therapies does not present such a scenario. Poon et al. [23]
found that the distribution of nonsynonymous substitutions
along the gene is shaped by selection to PI resistance.
Moreover, antiviral therapies promote complex drug-specific
residue-residue interaction networks [23, 70, 71] that can
drive the coevolution of primary and secondary resistance
mutations [8, 23].

5. Genetic Impact of Diverse PIs on HIV-1

PR-Coding Genes: A Computational Study
The HIV-1 PR is one of the most used drug targets for combating HIV with a number of chemically diverse inhibitors
that have already been tested [72, 73]. This section includes a
computational analysis of nucleotide diversity and molecular
adaptation of the PR-coding gene evolution under different
PIs.
5.1. Sample Collection. Samples of coding DNA sequences
that encode the HIV-1 PR (Pol region, subtype B) were
collected from the Stanford HIV Drug Resistance Database
[74, 75]. Subtype B was used because most (99%) of datasets
available in the database belong to this subtype and there is
not enough data to analyze other subtypes. For each HIV-1
patient, a clonal sequence was collected under no-treatment
and another one was collected after a particular treatment
based on a single PI or a PIs combination. According to
the detailed information provided by the database [74, 75],
the patients did not receive other treatments. Therefore, to
study each treatment (hereafter, evolutionary scenario) two
datasets (pool of sequences before and after treatment) were

3
obtained. In particular, for each evolutionary scenario, a
dataset includes coding sequences collected before a given
treatment and the other dataset includes coding sequences
collected after such a treatment, and both datasets come from
the same patients. As suggested by Kosakovsky Pond and
Frost [76], scenarios with sample size lower than 10 were
not considered to avoid lack of power in the evolutionary
analysis (datasets with higher sample size can generate
accurate estimates of genetic diversity and nonsynonymous
to synonymous substitution rates ratio (dN/dS) [76]; see also
[3, 77]). A total of 13 evolutionary scenarios, all the currently
available scenarios from the database, were analyzed. Namely,
a control scenario (no-treatment in both datasets, scenario
1, 1011 patients) and scenarios with the following treatments:
amprenavir (APV, scenario 2, 15 patients), atazanavir (ATV,
scenario 3, 23 patients), indinavir (IDV, scenario 4, 77
patients), lopinavir (LPV, scenario 5, 34 patients), nelfinavir
(NFV, scenario 6, 317 patients), ritonavir (RTV, scenario 7,
24 patients), saquinavir (SQV, scenario 8, 35 patients), and
the PI combinations: IDV + RTV (scenario 9, 10 patients),
RTV + SQV (scenario 10, 11 patients), and IDV + RTV + SQV
(scenario 11, 11 patients). Two additional scenarios were also
studied where patients treated with IDV are then treated with
IDV + NFV (scenario 12, 13 patients) or IDV + RTV (scenario
13, 16 patients).
5.2. Analysis of Genetic Diversity and Recombination. Several
genetic statistics were applied to study the influence of PIs on
the genetic diversity of the PR-coding gene. (i) The overall
sequences divergence was computed with MEGA 6.0 [78].
(ii) Nucleotide diversity () was estimated by using the
pairwise nucleotide differences per site [79]. These metrics
considered indels as missing data. (iii) The genetic distance
between the two datasets of each evolutionary scenario was
computed by the Kullback-Leibler (KL) divergence [80] and
considering indels as missing data and as an additional state.
This distance provides a comparative analysis of nucleotide
diversity distributions across sites between two datasets [81].
Briefly, the results show that almost all PIs lead to higher
levels of sequences divergence, pairwise nucleotide diversity,
and nucleotide diversity distribution across sites. Except
for LPV and NFV, all PIs increased the overall difference
between sequences (Figure 1(a)). Similar results are derived
from the estimates of nucleotide diversity although here only
LPV presented low levels of nucleotide diversity variation
(see Figure S1 in Supplementary Material available online at
http://dx.doi.org/10.1155/2015/395826). The highest levels of
diversity were generated from treatment with APV, IDV, and,
especially, PIs combinations. However, the increase of diversity could be caused by the emergence of resistance mutations
but also by mutations derived from the natural evolution of
the gene. Therefore, it is interesting to evaluate the correlation
between the variation of diversity and the corresponding
time period between samples. Figure 2(a) suggests that
there is no correlation between these parameters, which is
supported by a low correlation coefficient ( = 0.056). For
example, the control dataset (no-treatment) does not present
increase of diversity despite its long time period (11 months),
whereas treatment with APV generated one of the highest

Computational and Mathematical Methods in Medicine

Sequence difference variation (%)

2.0
1.5
1.0
0.5
0
0.5

1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0

/APV r = 0.016
/IDV r = 0.060 /IDV, RTV, SQV
/RTV, SQV r = 0.258

2.0
1.5

/IDV, RTV r = 0.189

/RTV
IDV/IDV, NFV r = 0.034
r = 0.507
IDV/IDV, RTV r = 0.018
/ATV
r = 0.0004
/SQV r = 0.003

1.0
0.5

/NFV
r = 0.001

0.0
0.5

r = 0.084

r = 0.056

/ r = 0.0003
/LPV r = 0.069

10

20

30

40

50

60

Time period (months)

levels of diversity in only 4 months (Figure 2). In addition,

correlation coefficients within scenarios (among patients
from a particular scenario) were also very low, most of them
under 0.1 (Figure 2(a)). A normalization dividing the genetic
diversity gradient by the time period between samples also
indicated the increase of genetic diversity with most of PIs
(Figure 2(b)). However, the normalization must be carefully
interpreted because a longer time period does not necessarily
lead to more diversity [82], which is actually indicated by the
described lack of correlation.
The analysis of nucleotide diversity distribution across
sites between the two alignments showed similar findings
for most of PIs (Figure 1(b)). Notice that this nucleotide
distribution can be more influenced by several mutations
at specific positions [80] and therefore this statistic might
be more sensible to detect resistance mutations. The results
show an influence of all PIs on the nucleotide diversity distribution (Figure 1(b)), although this influence varies among
inhibitors. Again, the long KL distance derived from drug
combination therapies is remarkable.

0.2
0.1

/SQV

/RTV

/NFV

/IDV

0.0
0.1

/IDV,
RTV
/RTV,
SQV
/IDV, RTV,
SQV
IDV/IDV,
NFV
IDV/IDV,
RTV

Figure 1: Overall sequence diversity variation and Kullback-Leibler

divergence. (a) Variation of overall sequence difference between
the two datasets of each evolutionary scenario (after treatment
before treatment ). Indels are considered as missing data. Error bars
indicate standard error. Reference values are shown in Table S1 (Supplementary Material). (b): Kullback-Leibler distance, nucleotide
diversity distribution, between the two datasets of each evolutionary
scenario. Dark grey bars consider indels as a new state whereas
clear grey bars consider indels as missing data. Error bars indicate
standard error across sites. indicates nave-treatment patients.

0.3

/LPV

(b)

0.4

/ATV

Indels as a new state

Indels as missing data

0.5

/APV

Evolutionary scenario

0.6

/IDV,
RTV
/RTV,
SQV
/IDV, RTV,
SQV
IDV/IDV,
NFV
IDV/IDV,
RTV

/SQV

/RTV

/NFV

/LPV

/IDV

/APV

/ATV

(a)

Kullback-Leibler
distance (%)

(a)

2.5

Sequence difference variation/

time period

Sequence difference
variation (%)

2.5

Evolutionary scenario
(b)

Figure 2: Sequence difference variation as a function of time

interval between samples. (a) Variation of the overall sequence
difference between the two datasets of each evolutionary scenario (after treatment before treatment ) is represented in the -axis
(mean and standard error). The time period between both samples
(after treatment before treatment ) is represented in the -axis (mean
and standard error from all patients of the scenario). The correlation
coefficient between both parameters among all the scenarios is =
0.056, suggesting absence of correlation. Correlation coefficients
within each evolutionary scenario (among patients) between these
parameters are also shown in the plot and ranges from 0.0003 to
0.507, although most of them are under 0.1. (b) Genetic diversity
gradient divided by the corresponding time period. Error bars
indicate standard error. indicates nave-treatment patients.
Reference values are shown in Table S1 (Supplementary Material).

Absence of recombination breakpoints was found with

the single breakpoint position (SBP) method [83], implemented in the Hyphy package [84], and with the recombination detection methods implemented in the RDP framework
[85].
5.3. Signatures of Molecular Adaptation. The best-fit model
of DNA substitution was selected with jModelTest [86] under
the Bayesian information criterion (BIC), as suggested by
[87]. Then, maximum likelihood (ML) phylogenetic trees
were inferred with PhyML [88] under the corresponding
substitution model. These trees were used to perform the
molecular adaptation inferences. Estimates of dN/dS at both
global (sequence) and local (codon) levels were performed

/IDV,
RTV
/RTV,
SQV
/IDV, RTV,
SQV
IDV/IDV,
NFV
IDV/IDV,
RTV

/SQV

/RTV

/NFV

/LPV

/IDV

/ATV

/APV

0.35
0.30
0.25
0.20
0.15
0.10
0.05
0
0.05

Global dN/dS variation

Computational and Mathematical Methods in Medicine

Evolutionary scenario

Figure 3: Global nonsynonymous to synonymous substitution rates

ratio (dN/dS) variation. Variation of global dN/dS between the
two datasets of each evolutionary scenario (/after treatment
/before treatment ). Error bars indicate 95% CI. indicates
nave-treatment patients. Reference values are shown in Table S3
(Supplementary Material).

with the Fixed Effects Likelihood (FEL) method [76] implemented in the Hyphy package. Notice that this ML-based
method provides very accurate estimates [76] and it is
commonly used in population genetics and virus evolution
(e.g., [3]).
Figure 3 shows the variation of global dN/dS estimates
from datasets collected before and after a treatment. All the
PIs promoted increased estimates of dN/dS, especially when
the treatment is based on PIs combination. By contrast, in
absence of treatment the estimated dN/dS declined with time.
At the local level, almost all the PIs promoted an increase
of significant ( value < 0.05) PSSs along the PR-coding
sequence (Figure S2 and Table S4, Supplementary Material).
In general, a large number of NSSs were detected in all
datasets (Table S4) without showing a clear relationship with
the presence or absence of treatment. All these results are
discussed in the next section.

6. Concluding Remarks
The fast population range contractions and fragmentation
produced during the therapy can reduce the overall diversity
of viral strains [29] and, by contrast, the emergence of
resistance mutations caused from the rapid evolution of HIV
allows preserving or increasing the levels of nucleotide diversity of viral protein-coding genes of the drug target. Indeed,
resistance mutations can be rare, but also recurrent enough
until they reach resistance, and can generate positive selection
driving the fixation of favorable viral strains [23, 67]. At this
level two opposite selective pressures seem to act. While most
of sites evolve under strong purifying selection, probably
caused by the hosts immune system and the therapy, other
sites evolve under diversifying selection, probably caused
by the viral molecular adaptation to the new environment
established by the therapy, and can present complex residueresidue interaction networks suggesting dependent evolution
among sites [23, 70].

5
The evolutionary analysis included in this work supported such considerations. It also showed that different PIs
can promote different influences on the genetic diversity
of the viral PR-coding gene. For example, IDV and APV
induced the highest levels of diversity (among treatments
with a single PI) and PIs combination induced very high
levels of genetic diversity, especially when 3 PIs are applied
jointly. As noted, this increase of diversity can be related with
the emergence of resistance mutations [33]. The reason why
some inhibitors induce more diversity than others requires
complex structural analysis of enzyme-inhibitor interactions,
which is an important topic of research [89, 90]. On the other
hand, absence of recombination breakpoints was found in
the analyzed PR-coding genes. This could be caused by limitations to detect recombination under low genetic diversity
levels [59] or just because recombination was not required to
generate drug resistance. Indeed, recombination could occur
in other genomic regions [63] (i.e., recombinants with breakpoints in Gag and Pol may present selection against [63, 91]).
Concerning molecular adaptation, the results showed that the
wild evolution of the virus presents an overall decrease of the
global dN/dS, without PSSs and where most of sites evolved
under significant purifying selection. This is probably caused
by the sharp purifying selection induced by the hosts immune
system. On the other hand, PIs often promoted an overall
increase of the global dN/dS (see [23, 68] and Figure 3), which
is most of times accompanied by the emergence of significant
PSSs along the gene (see [23] and Figure S2). As expected,
the largest dN/dS increase occurs under treatments with PIs
combination (see [23] and Figure 3). These signatures of
molecular adaptation are related with the amount of genetic
diversity induced by the PIs and indicate the primary importance of adaptation in the evolutionary process of the PRcoding gene under PIs. On the other hand, a large number of
codons evolved under negative (purifying) selection, which
indicates the presence of strong selective pressures, as noted
probably caused by the hosts immune system and the therapy.
Understanding molecular evolution of the virus can help
us develop more realistic models of HIV evolution [23, 69,
71], correlate the disease progression with the evolution of
the viral population [28, 92], and predict resistance (i.e., by
genotypic-resistance testing [24, 25, 93, 94]) and common
ancestry [18], or vaccine design [1719]. Nevertheless, HIV1 evolution is complex and other phenomena should also be
taken into account as much as possible in the models, for
example, different hosts immune responses, clinical stage,
HIV-1 compartmentalization [95], or infection with multiple
viral variants, although the latter presents an overall low
incidence [53, 96].
Since antiviral therapies affect genetic diversity of the
virus by strong selective pressures, models of HIV-1 evolution
should accommodate such effects in order to mimic these
scenarios for purposes such as robust genotypic resistance
testing and treatments design. Importantly, models of HIV-1
evolution should be as realistic as possible in order to provide
accurate predictions. A possibility could be the consideration
of a fitness landscape (e.g., [97]) to develop parametric
models. However, the design and computation of a realistic
fitness function are too convoluted due to complex processes

6
that affect viral genetic diversity such as antiviral therapies (as
noted in this paper). An easier, but less robust, alternative can
be the development of scenario-specific empirical models.
As shown in this paper, different therapies must be modeled
with different models of evolution since different therapies
can promote different genetic consequences in the virus.
Much more research is needed (i.e., the consideration of
associations between observed genotypes and phenotypic
resistance in models of HIV-1 evolution) but my impression
is that HIV-1 therapies will benefit from more consideration
of evolutionary information.

Conflict of Interests
The author declares that there is no conflict of interests
regarding the publication of this paper.

Acknowledgments
The author wants to thank three anonymous reviewers for
insightful comments. This work was supported by the Spanish
Government through the Juan de la Cierva Fellowship JCI2011-10452 and by the Portuguese Government through the
FCT Starting Grant IF/00955/2014.

References
[1] UNAIDS, 2013 report on the global AIDS epidemic, in
Joint United Nations Programme on HIV/AIDS, World Health
Organization, 2013.
[2] B. S. Peters and K. Conway, Therapy for HIV: past, present, and
future, Advances in Dental Research, vol. 23, no. 1, pp. 2327,
2011.
[3] M. Perez-Losada, D. Posada, M. Arenas et al., Ethnic differences in the adaptation rate of HIV gp120 from a vaccine trial,
Retrovirology, vol. 6, article 67, 2009.
[4] A. McMichael, L. J. Picker, J. P. Moore, and D. R. Burton,
Another HIV vaccine failure: where to next? Nature medicine,
vol. 19, no. 12, pp. 15761577, 2013.
[5] K. S. Slobod, M. Bonsignori, S. A. Brown, X. Zhan, J. Stambas,
and J. L. Hurwitz, HIV vaccines: brief review and discussion
of future directions, Expert Review of Vaccines, vol. 4, no. 3, pp.
305313, 2005.
[6] B. T. Korber, N. L. Letvin, and B. F. Haynes, T-cell vaccine
strategies for human immunodeficiency virus, the virus with a
thousand faces, Journal of Virology, vol. 83, no. 17, pp. 8300
8314, 2009.
[7] J. D. Siliciano and R. F. Siliciano, The latent reservoir for HIV-1
in resting CD4+ T cells: a barrier to cure, Current Opinion in
HIV and AIDS, vol. 1, no. 2, pp. 121128, 2006.
[8] L. Menendez-Arias, Molecular basis of human immunodeficiency virus type 1 drug resistance: overview and recent
developments, Antiviral Research, vol. 98, no. 1, pp. 93120,
2013.
[9] A. M. Wensing, V. Calvez, H. F. Gunthard et al., 2014 Update
of the drug resistance mutations in HIV-1, Topics in Antiviral
Medicine, vol. 22, no. 3, pp. 642650, 2014.
[10] D. Shriner, A. G. Rodrigo, D. C. Nickle, and J. I. Mullins,
Pervasive genomic recombination of HIV-1 in vivo, Genetics,
vol. 167, no. 4, pp. 15731583, 2004.

Computational and Mathematical Methods in Medicine

[11] M. Gotte, The distinct contributions of fitness and genetic
barrier to the development of antiviral drug resistance, Current
Opinion in Virology, vol. 2, no. 5, pp. 644650, 2012.
[12] L. Menendez-Arias, M. A. Martnez, M. E. Quinones-Mateu,
and J. Martinez-Picado, Fitness variations and their impact on
the evolution of antiretroviral drug resistance, Current Drug
Targets: Infectious Disorders, vol. 3, no. 4, pp. 355371, 2003.
[13] R. van Leeuwen, C. Katlama, R. L. Murphy et al., A randomized
trial to study first-line combination therapy with or without a
protease inhibitor in HIV-1-infected patients, AIDS, vol. 17, no.
7, pp. 987999, 2003.
[14] K. Hertogs, S. Bloor, S. D. Kemp et al., Phenotypic and
genotypic analysis of clinical HIV-1 isolates reveals extensive
protease inhibitor cross-resistance: a survey of over 6000
samples, AIDS, vol. 14, no. 9, pp. 12031210, 2000.
[15] N. M. King, L. Melnick, M. Prabu-Jeyabalan et al., Lack of
synergy for inhibitors targeting a multi-drug-resistant HIV-1
protease, Protein Science, vol. 11, no. 2, pp. 418429, 2002.
[16] E. Castro-Nallar, M. Perez-Losada, G. F. Burton, and K. A.
Crandall, The evolution of HIV: inferences using phylogenetics, Molecular Phylogenetics and Evolution, vol. 62, no. 2, pp.
777792, 2012.
[17] M. Arenas and D. Posada, Computational design of centralized
HIV-1 genes, Current HIV Research, vol. 8, no. 8, pp. 613621,
2010.
[18] N. A. Dona-Rose, G. H. Learn, A. G. Rodrigo et al., Human
immunodeficiency virus type 1 subtype B ancestral envelope
protein is functional and elicits neutralizing antibodies in
rabbits similar to those elicited by a circulating subtype B
envelope, Journal of Virology, vol. 79, no. 17, pp. 1121411224,
2005.
[19] D. C. Nickle, M. A. Jensen, G. S. Gottlieb et al., Consensus
and ancestral state HIV vaccines, Science, vol. 299, no. 5612, pp.
15151518, 2003.
[20] A. R. Lemmon and E. C. Moriarty, The importance of proper
model assumption in bayesian phylogenetics, Systematic Biology, vol. 53, no. 2, pp. 265277, 2004.
[21] D. C. Nickle, L. Heath, M. A. Jensen, P. B. Gilbert, J. I. Mullins,
and S. L. K. Pond, HIV-specific probabilistic models of protein
evolution, PLoS ONE, vol. 2, no. 6, article e503, 2007.
[22] C. Fraser, K. Lythgoe, G. E. Leventhal et al., Virulence and
pathogenesis of HIV-1 infection: an evolutionary perspective,
Science, vol. 343, no. 6177, Article ID 1243727, 2014.
[23] A. F. Y. Poon, S. L. Kosakovsky Pond, D. D. Richman, and
S. D. W. Frost, Mapping protease inhibitor resistance to
human immunodeficiency virus type 1 sequence polymorphisms within patients, Journal of Virology, vol. 81, no. 24, pp.
1359813607, 2007.
[24] A. R. Zolopa, R. W. Shafer, A. Warford et al., HIV-1 genotypic
resistance patterns predict response to saquinavir-ritonavir
therapy in patients in whom previous protease inhibitor therapy
had failed, Annals of Internal Medicine, vol. 131, no. 11, pp. 813
821, 1999.
[25] J. Durant, P. Clevenbergh, P. Halfon et al., Drug-resistance
genotyping in HIV-1 therapy: the VIRADAPT randomised
controlled trial, The Lancet, vol. 353, no. 9171, pp. 21952199,
1999.
[26] L. Ercoli, L. Sarmati, E. Nicastri et al., HIV phenotype switching during antiretroviral therapy: emergence of saquinavirresistant strains with less cytopathogenicity, AIDS, vol. 11, no.
10, pp. 12111217, 1997.

Computational and Mathematical Methods in Medicine

[27] S. Harada, K. Yoshimura, A. Yamaguchi, S. Boonchawalit, K.
Yusa, and S. Matsushita, Impact of antiretroviral pressure on
selection of primary human immunodeficiency virus type 1
envelope sequences in vitro, Journal of General Virology, vol.
94, part 5, pp. 933943, 2013.
[28] R. M. Troyer, K. R. Collins, A. Abraha et al., Changes in human
immunodeficiency virus type 1 fitness and genetic diversity
during disease progression, Journal of Virology, vol. 79, no. 14,
pp. 90069018, 2005.
[29] M. Arenas, N. Ray, M. Currat, and L. Excoffier, Consequences
of range contractions and range shifts on molecular diversity,
Molecular Biology and Evolution, vol. 29, no. 1, pp. 207218, 2012.
[30] S. Mona, N. Ray, M. Arenas, and L. Excoffier, Genetic consequences of habitat fragmentation during a range expansion,
Heredity, vol. 112, no. 3, pp. 291299, 2014.
[31] K. M. Kitrinos, J. A. E. Nelson, W. Resch, and R. Swanstrom,
Effect of a protease inhibitor-induced genetic bottleneck on
human immunodeficiency virus type 1 env gene populations,
Journal of Virology, vol. 79, no. 16, pp. 1062710637, 2005.
[32] A. Iban ez, B. Clotet, and M.-A. Martnez, Human immunodeficiency virus type 1 population bottleneck during indinavir
therapy causes a genetic drift in the env quasispecies, Journal
of General Virology, vol. 81, no. 1, pp. 8595, 2000.
[33] S. A. Rabi, G. M. Laird, C. M. Durand et al., Multi-step inhibition explains HIV-1 protease inhibitor pharmacodynamics and
resistance, Journal of Clinical Investigation, vol. 123, no. 9, pp.
38483860, 2013.
[34] P. Zhong, Q. Pan, Z. Ning et al., Genetic diversity and drug
resistance of human immunodeficiency virus type 1 (HIV-1)
strains circulating in Shanghai, AIDS Research and Human
Retroviruses, vol. 23, no. 7, pp. 847856, 2007.
[35] I. Chen, L. Khaki, J. C. Lindsey et al., Association of pol
diversity with antiretroviral treatment outcomes among HIVinfected african children, PLoS ONE, vol. 8, no. 11, Article ID
e81213, 2013.
[36] T. D. Wu, C. A. Schiffer, M. J. Gonzales et al., Mutation
patterns and structural correlates in human immunodeficiency
virus type 1 protease following different protease inhibitor
treatments, Journal of Virology, vol. 77, no. 8, pp. 48364847,
2003.
[37] R. W. Shafer, R. Kantor, and M. J. Gonzales, The genetic
basis of HIV-1 resistance to reverse transcriptase and protease
inhibitors, AIDS Reviews, vol. 2, no. 4, pp. 211228, 2000.
[38] D. J. de Sa-Filho, M. D. S. Soares, V. Candido et al., HIV type 1
pol gene diversity and antiretroviral drug resistance mutations
in Santos, Brazil, AIDS Research and Human Retroviruses, vol.
24, no. 3, pp. 347353, 2008.
[39] R. Kantor and D. Katzenstein, Polymorphism in HIV-1 nonsubtype b protease and reverse transcriptase and its potential
impact on drug susceptibility and drug resistance evolution,
AIDS Reviews, vol. 5, no. 1, pp. 2535, 2003.
[40] L. B. Arruda, L. A. M. Fonseca, A. J. S. Duarte, and J. Casseb,
Genetic diversity on the integrase region of the pol gene among
HIV type 1-infected patients naive for integrase inhibitors in Sao
Paulo City, Brazil, AIDS Research and Human Retroviruses, vol.
26, no. 1, pp. 105107, 2010.
[41] G. S. Gottlieb, R. A. Smith, N. M. Dia Badiane et al., Hiv2 integrase variation in integrase inhibitor-nave adults in
Senegal, West Africa, PLoS ONE, vol. 6, no. 7, Article ID e22204,
2011.
[42] C. Garrido, A. M. Geretti, N. Zahonero et al., Integrase
variability and susceptibility to HIV integrase inhibitors: impact

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]

[51]

[52]

[53]

[54]

[55]

[56]

[57]

of subtypes, antiretroviral experience and duration of HIV

infection, The Journal of Antimicrobial Chemotherapy, vol. 65,
no. 2, pp. 320326, 2010.
A. Piralla, S. Paolucci, R. Gulminetti, G. Comolli, and F.
Baldanti, HIV integrase variability and genetic barrier in
antiretroviral nave and experienced patients, Virology Journal,
vol. 8, article 149, 2011.
R. Kulkarni, R. Hluhanich, D. M. McColl, M. D. Miller, and K.
L. White, The combined anti-HIV-1 activities of emtricitabine
and tenofovir plus the integrase inhibitor elvitegravir or raltegravir show high levels of synergy in vitro, Antimicrobial Agents
and Chemotherapy, vol. 58, no. 10, pp. 61456150, 2014.
R. Najera, E. Delgado, L. Perez-Alvarez, and M. M. Thomson,
Genetic recombination and its role in the development of the
HIV-1 pandemic, AIDS, vol. 16, supplement 4, pp. S3S16, 2002.
D. Moradigaravand, R. Kouyos, T. Hinkley et al., Recombination accelerates adaptation on a large-scale empirical fitness
landscape in HIV-1, PLoS Genetics, vol. 10, no. 6, Article ID
e1004439, 2014.
M. Perez-Losada, M. Arenas, J. C. Galan, F. Palero, and F.
Gonzalez-Candelas, Recombination in viruses: mechanisms,
methods of study, and evolutionary consequences, Infection,
Genetics and Evolution, vol. 30, pp. 296307, 2015.
M. Anisimova, R. Nielsen, and Z. Yang, Effect of recombination on the accuracy of the likelihood method for detecting
positive selection at amino acid sites, Genetics, vol. 164, no. 3,
pp. 12291236, 2003.
M. Arenas and D. Posada, The influence of recombination on
the estimation of selection from coding sequence alignments,
in Natural Selection: Methods and Applications, M. A. Fares, Ed.,
pp. 112125, CRC Press/Taylor & Francis, Boca Raton, Fla, USA,
2014.
M. H. Schierup and J. Hein, Consequences of recombination
on traditional phylogenetic analysis, Genetics, vol. 156, no. 2,
pp. 879891, 2000.
M. Arenas and D. Posada, The effect of recombination on the
reconstruction of ancestral sequences, Genetics, vol. 184, no. 4,
pp. 11331139, 2010.
M. T. Bretscher, C. L. Althaus, V. Muller, and S. Bonhoeffer,
Recombination in HIV and the evolution of drug resistance:
for better or for worse? BioEssays, vol. 26, no. 2, pp. 180188,
2004.
M. Pernas, C. Casado, R. Fuentes, M. J. Perez-Elas, and C.
Lopez-Galndez, A dual superinfection and recombination
within HIV-1 subtype B 12 years after primoinfection, Journal of
Acquired Immune Deficiency Syndromes, vol. 42, no. 1, pp. 1218,
2006.
C. Fraser, HIV recombination: what is the impact on antiretroviral therapy? Journal of the Royal Society Interface, vol. 2, no.
5, pp. 489503, 2005.
N. N. V. Vijay, R. Ajmani, A. S. Perelson, and N. M. Dixit,
Recombination increases human immunodeficiency virus fitness, but not necessarily diversity, Journal of General Virology,
vol. 89, no. 6, pp. 14671477, 2008.
D. Shriner, R. Shankarappa, M. A. Jensen et al., Influence of
random genetic drift on human immunodeficiency virus type 1
env evolution during chronic infection, Genetics, vol. 166, no.
3, pp. 11551164, 2004.
L. J. Costa, P. Munerato, R. S. Diaz, and A. Tanuri, Generation
of intersubtype human immunodeficiency virus type 1 recombinants in env gene in vitro: influences in the biological behavior

[58]

[59]

[60]

[61]

[62]

[63]

[64]

[65]

[66]

[67]

[68]

[69]

[70]

[71]

[72]

Computational and Mathematical Methods in Medicine

and in the establishment of productive infections, Virology, vol.
268, no. 2, pp. 440451, 2000.
M. E. Quinones-Mateu, Y. Gao, S. C. Ball, A. J. Marozsan, A.
Abraha, and E. J. Arts, In vitro intersubtype recombinants of
human immunodeficiency virus type 1: comparison to recent
and circulating in vivo recombinant forms, Journal of Virology,
vol. 76, no. 19, pp. 96009613, 2002.
D. Posada and K. A. Crandall, Evaluation of methods for
detecting recombination from DNA sequences: computer simulations, Proceedings of the National Academy of Sciences of the
United States of America, vol. 98, no. 24, pp. 1375713762, 2001.
A. Carvajal-Rodrguez, K. A. Crandall, and D. Posada, Recombination favors the evolution of drug resistance in HIV-1 during
antiretroviral therapy, Infection, Genetics and Evolution, vol. 7,
no. 4, pp. 476483, 2007.
B. Shi, C. Kitchen, B. Weiser et al., Evolution and recombination of genes encoding HIV-1 drug resistance and tropism
during antiretroviral therapy, Virology, vol. 404, no. 1, pp. 520,
2010.
C. L. Althaus and S. Bonhoeffer, Stochastic interplay between
mutation and recombination during the acquisition of drug
resistance mutations in human immunodeficiency virus type 1,
Journal of Virology, vol. 79, no. 21, pp. 1357213578, 2005.
J. Archer, J. W. Pinney, J. Fan et al., Identifying the important
HIV-1 recombination breakpoints, PLoS Computational Biology, vol. 4, no. 9, Article ID e1000178, 7 pages, 2008.
M. Nijhuis, C. A. B. Boucher, P. Schipper, T. Leitner, R. Schuurman, and J. Albert, Stochastic processes strongly influence
HIV-1 evolution during suboptimal protease-inhibitor therapy,
Proceedings of the National Academy of Sciences of the United
States of America, vol. 95, no. 24, pp. 1444114446, 1998.
Y. Kim and H. A. Orr, Adaptation in sexuals vs. asexuals: clonal
interference and the Fisher-Muller model, Genetics, vol. 171, no.
3, pp. 13771386, 2005.
M. Foll, Y.-P. Poh, N. Renzette et al., Influenza virus drug
resistance: a time-sampled population genetics perspective,
PLoS Genetics, vol. 10, no. 2, Article ID e1004185, 2014.
A. F. Y. Poon, S. L. Kosakovsky Pond, P. Bennett, D. D. Richman,
A. J. Leigh Brown, and S. D. W. Frost, Adaptation to human
populations is revealed by within-host polymorphisms in HIV1 and hepatitis C virus, PLoS Pathogens, vol. 3, no. 3, article e45,
2007.
C. Pan, J. Kim, L. Chen, Q. Wang, and C. Lee, The HIV positive
selection mutation database, Nucleic Acids Research, vol. 35,
supplement 1, pp. D371D375, 2007.
J. Woo, D. L. Robertson, and S. C. Lovell, Constraints on HIV1 diversity from protein structure, Journal of Virology, vol. 84,
no. 24, pp. 1299513003, 2010.
M. John, C. B. Moore, I. R. James, and S. A. Mallal, Interactive
selective pressures of HLA-restricted immune responses and
antiretroviral drugs on HIV-1, Antiviral Therapy, vol. 10, no. 4,
pp. 551555, 2005.
N. Beerenwinkel, B. Schmidt, H. Walter et al., Diversity and
complexity of HIV-1 drug resistance: a bioinformatics approach
to predicting phenotype from genotype, Proceedings of the
National Academy of Sciences of the United States of America,
vol. 99, no. 12, pp. 82718276, 2002.
M. Arenas, M. C. Villaverde, and F. Sussman, Prediction
and analysis of binding affinities for chemically diverse HIV1 PR inhibitors by the modified SAFE p approach, Journal of
Computational Chemistry, vol. 30, no. 8, pp. 12291240, 2009.

[73] A. M. J. Wensing, N. M. van Maarseveen, and M. Nijhuis,

Fifteen years of HIV Protease Inhibitors: raising the barrier to
resistance, Antiviral Research, vol. 85, no. 1, pp. 5974, 2010.
[74] S. Y. Rhee, M. J. Gonzales, R. Kantor, B. J. Betts, J. Ravela,
and R. W. Shafer, Human immunodeficiency virus reverse
transcriptase and protease sequence database, Nucleic Acids
Research, vol. 31, no. 1, pp. 298303, 2003.
[75] R. W. Shafer, Rationale and uses of a public HIV drugresistance database, Journal of Infectious Diseases, vol. 194,
supplement 1, pp. S51S58, 2006.
[76] S. L. Kosakovsky Pond and S. D. W. Frost, Not so different
after all: a comparison of methods for detecting amino acid sites
under selection, Molecular Biology and Evolution, vol. 22, no. 5,
pp. 12081222, 2005.
[77] R. Batorsky, M. F. Kearney, S. E. Palmer, F. Maldarelli, I. M.
Rouzine, and J. M. Coffin, Estimate of effective recombination
rate and average selection coefficient for HIV in chronic
infection, Proceedings of the National Academy of Sciences of the
United States of America, vol. 108, no. 14, pp. 56615666, 2011.
[78] K. Tamura, G. Stecher, D. Peterson, A. Filipski, and S. Kumar,
MEGA6: Molecular evolutionary genetics analysis version 6.0,
Molecular Biology and Evolution, vol. 30, no. 12, pp. 27252729,
2013.
[79] M. Nei, Molecular Evolutionary Genetics, Columbia University
Press, New York, NY, USA, 1987.
[80] S. Kullback, Letter to the editor: the KullbackLeibler distance, The American Statistician, vol. 41, no. 4, pp. 340341,
1987.
[81] M. Arenas, H. G. Dos Santos, D. Posada, and U. Bastolla, Protein evolution along phylogenetic histories under structurally
constrained substitution models, Bioinformatics, vol. 29, no. 23,
pp. 30203028, 2013.
[82] G. Bello, C. Casado, S. Garca et al., Lack of temporal structure
in the short term HIV-1 evolution within asymptomatic nave
patients, Virology, vol. 362, no. 2, pp. 294303, 2007.
[83] S. L. K. Pond, D. Posada, M. B. Gravenor, C. H. Woelk, and S. D.
W. Frost, Automated phylogenetic detection of recombination
using a genetic algorithm, Molecular Biology and Evolution, vol.
23, no. 10, pp. 18911901, 2006.
[84] S. L. Kosakovsky Pond, S. D. W. Frost, and S. V. Muse, HyPhy:
hypothesis testing using phylogenies, Bioinformatics, vol. 21,
no. 5, pp. 676679, 2005.
[85] D. P. Martin, P. Lemey, M. Lott, V. Moulton, D. Posada, and
P. Lefeuvre, RDP3: a flexible and fast computer program for
analyzing recombination, Bioinformatics, vol. 26, no. 19, Article
ID btq467, pp. 24622463, 2010.
[86] D. Posada, jModelTest: phylogenetic model averaging, Molecular Biology and Evolution, vol. 25, no. 7, pp. 12531256, 2008.
[87] A. Luo, H. Qiao, Y. Zhang et al., Performance of criteria for
selecting evolutionary models in phylogenetics: a comprehensive study based on simulated datasets, BMC Evolutionary
Biology, vol. 10, no. 1, article 242, 2010.
[88] S. Guindon, J.-F. Dufayard, V. Lefort, M. Anisimova, W. Hordijk,
and O. Gascuel, New algorithms and methods to estimate
maximum-likelihood phylogenies: assessing the performance
of PhyML 3.0, Systematic Biology, vol. 59, no. 3, pp. 307321,
2010.
[89] T. Hinkley, J. Martins, C. Chappey et al., A systems analysis of
mutational effects in HIV-1 protease and reverse transcriptase,
Nature Genetics, vol. 43, no. 5, pp. 487490, 2011.

Computational and Mathematical Methods in Medicine

[90] G. J. Henderson, S.-K. Lee, D. M. Irlbeck et al., Interplay
between single resistance-associated mutations in the HIV-1
protease and viral infectivity, protease activity, and inhibitor
sensitivity, Antimicrobial Agents and Chemotherapy, vol. 56, no.
2, pp. 623633, 2012.
[91] D. P. Martin, E. van der Walt, D. Posada, and E. P. Rybicki, The
evolutionary value of recombination is constrained by genome
modularity, PLoS genetics, vol. 1, no. 4, p. e51, 2005.
[92] M. A. Nowak, R. M. Anderson, M. C. Boerlijst, S. Bonhoeffer,
R. M. May, and A. J. McMichael, HIV-1 evolution and disease
progression, Science, vol. 274, no. 5289, pp. 10081011, 1996.
[93] B. Zollner, H.-H. Feucht, L. Weitner et al., Application of HIV1 genotypic-resistance testing prevents the evolution of further
resistance mutations in heavily pretreated patients, Journal of
Clinical Virology, vol. 21, no. 1, pp. 3745, 2001.
[94] M. L. Branham, E. A. Ross, and T. Govender, Predictive models
for maximum recommended therapeutic dose of antiretroviral
drugs, Computational and Mathematical Methods in Medicine,
vol. 2012, Article ID 469769, 9 pages, 2012.
[95] G. Tirado, G. Jove, R. Kumar et al., Compartmentalization of
drug resistance-associated mutations in a treatment-naive HIVinfected female, AIDS Research and Human Retroviruses, vol.
20, no. 6, pp. 684686, 2004.
[96] D. M. Smith, J. K. Wong, G. K. Hightower et al., Incidence of
HIV superinfection following primary infection, Journal of the
American Medical Association, vol. 292, no. 10, pp. 11771178,
2004.
[97] R. Lorenzo-Redondo, S. Delgado, F. Moran, and C. LopezGalindez, Realistic three dimensional fitness landscapes generated by self organizing maps for the analysis of experimental
HIV-1 evolution, PLoS ONE, vol. 9, no. 2, Article ID e88579,
2014.