Analysis of HIV-1 pol sequences using Bayesian Networks: implications for drug resistance

Comparing mutational pathways to lopinavir resistance in HIV-1 subtypes B versus C

Although combination antiretroviral therapies seem to be effective at controlling HIV-1 infections regardless of the viral subtype, there is increasing evidence for subtype-specific drug resistance mutations. The order and rates at which resistance mutations accumulate in different subtypes also remain poorly understood. Most of this knowledge is derived from studies of subtype B genotypes, despite not being the most abundant subtype worldwide. Here, we present a methodology for the comparison of mutational networks in different HIV-1 subtypes, based on Hidden Conjunctive Bayesian Networks (H-CBN), a probabilistic model for inferring mutational networks from cross-sectional genotype data. We introduce a Monte Carlo sampling scheme for learning H-CBN models for a larger number of resistance mutations and develop a statistical test to assess differences in the inferred mutational networks between two groups. We apply this method to infer the temporal progression of mutations conferring resistance to the protease inhibitor lopinavir in a large cross-sectional cohort of HIV-1 subtype C genotypes from South Africa, as well as to a data set of subtype B genotypes obtained from the Stanford HIV Drug Resistance Database and the Swiss HIV Cohort Study. We find strong support for different initial mutational events in the protease, namely at residue 46 in subtype B and at residue 82 in subtype C. The inferred mutational networks for subtype B versus C are significantly different sharing only five constraints on the order of accumulating mutations with mutation at residue 54 as the parental event. The results also suggest that mutations can accumulate along various alternative paths within subtypes, as opposed to a unique total temporal ordering. Beyond HIV drug resistance, the statistical methodology is applicable more generally for the comparison of inferred mutational networks between any two groups.

There is a disparity in the distribution of infections by HIV-1 subtype in the world. Subtype B is predominant in America, Australia and western and central Europe, and most therapeutic strategies are based on research and clinical studies on this subtype. However, non-B subtypes represent the majority of global HIV-1 infections; e.g., subtype C alone accounts for nearly half of all HIV-1 infections. We present a statistical framework enabling the comparison of patterns of accumulating mutations in different HIV-1 subtypes. Specifically, we compare the temporal ordering of lopinavir resistance mutations in HIV-1 subtypes B versus C. To this end, we combine the Hidden Conjunctive Bayesian Network (H-CBN) model with an approximate inference scheme enabling comparisons of larger networks. We show that the development of resistance to lopinavir differs significantly between subtypes B and C, such that findings based on subtype B sequences can not always be applied to sybtype C. The described methodology is suitable for comparing different subgroups in the context of other evolutionary processes.

Gag-protease coevolution shapes the outcome of lopinavir-inclusive treatment regimens in chronically infected HIV-1 subtype C patients

MOTIVATION

Commonly, protease inhibitor failure is characterized by the development of multiple protease resistance mutations (PRMs). While the impact of PRMs on therapy failure are understood, the introduction of Gag mutations with protease remains largely unclear.

RESULTS

Here, we utilized phylogenetic analyses and Bayesian network learning as tools to understand Gag-protease coevolution and elucidate the pathways leading to Lopinavir failure in HIV-1 subtype C infected patients. Our analyses indicate that while PRMs coevolve in response to drug selection pressure within protease, the Gag mutations added to the existing network while specifically interacting with known Lopinavir failure PRMs. Additionally, the selection of mutations at specific positions in Gag-protease suggests that these coevolving mutational changes occurs to maintain structural integrity during Gag cleavage.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Complete genome sequence of human T-cell lymphotropic type 1 from patients with different clinical profiles, including infective dermatitis

The HTLV-1 is the first human retrovirus and is associated with several clinical syndromes, however, the pathogenesis of these clinical manifestations is still not fully understood. Furthermore, there are few complete genomes publicly available, about 0.12 complete genomes per 10,000 infected individuals and the databases have a major deficiency of sequences information. This study generated and characterized 31 HTLV-1 complete genomes sequences derived from individuals with Tropical Spastic Paraparesis/HTLV-1-Associated Myelopathy (TSP/HAM), Adult T-cell leukemia/lymphoma (ATL), infective dermatitis associated to HTLV-1 (IDH) and asymptomatic patients. These sequences are associated to clinical and epidemiological information about the patients. The sequencing data generated on Ion Torrent PGM platform were assembled and mapped against the reference HTLV-1 genome. These sequences were genotyped as Cosmopolitan subtype, Transcontinental subgroup. We identified the variants in the coding regions of the genome of the different clinical profiles, however, no statistical relation was detected. This study contributed to increase of HTLV-1 complete genomes in the world. Furthermore, to better investigate the contribution of HTLV-1 mutations for the disease outcome it is necessary to evaluate the interaction of the viral genome and characteristics of the human host.

Amino acid impact factor

Amino acid mutations in proteins are random and those mutations which are beneficial or neutral survive during the course of evolution. Conservation or co-evolution analyses are performed on the multiple sequence alignment of homologous proteins to understand how important different amino acids or groups of them are. However, these traditional analyses do not explore the directed influence of amino acid mutations, such as compensatory effects. In this work we develop a method to capture the directed evolutionary impact of one amino acid on all other amino acids, and provide a visual network representation for it. The method developed for these directed networks of inter- and intra-protein evolutionary interactions can also be used for noting the differences in amino acid evolution between the control and experimental groups. The analysis is illustrated with a few examples, where the method identifies several directed interactions of functionally critical amino acids. The impact of an amino acid is quantified as the number of amino acids that are influenced as a consequence of its mutation, and it is intended to summarize the compensatory mutations in large evolutionary sequence data sets as well as to rationally identify targets for mutagenesis when their functional significance can not be assessed using structure or conservation.

A genetic IFN/STAT1/FAS axis determines CD4 T stem cell memory levels and apoptosis in healthy controls and Adult T-cell Leukemia patients

Adult T-cell leukemia (ATL) is an aggressive, chemotherapy-resistant CD4+CD25+ leukemia caused by HTLV-1 infection, which usually develops in a minority of patients several decades after infection. IFN + AZT combination therapy has shown clinical benefit in ATL, although its mechanism of action remains unclear. We have previously shown that an IFN-responsive FAS promoter polymorphism in a STAT1 binding site (rs1800682) is associated to ATL susceptibility and survival. Recently, CD4 T stem cell memory (TSCM) Fashi cells have been identified as the hierarchical cellular apex of ATL, but a possible link between FAS, apoptosis, proliferation and IFN response in ATL has not been studied. In this study, we found significant ex vivo antiproliferative, antiviral and immunomodulatory effects of IFN-α treatment in short-term culture of primary mononuclear cells from ATL patients (n = 25). Bayesian Network analysis allowed us to integrate ex vivo IFN-α response with clinical, genetic and immunological data from ATL patients, thereby revealing a central role for FAS -670 polymorphism and apoptosis in the coordinated mechanism of action of IFN-α. FAS genotype-dependence of IFN-induced apoptosis was experimentally validated in an independent cohort of healthy controls (n = 20). The same FAS -670 polymorphism also determined CD4 TSCM levels in a genome-wide twin study (p = 7 × 10-11, n = 460), confirming a genetic link between apoptosis and TSCM levels. Transcriptomic analysis and cell type deconvolution confirmed the FAS genotype/TSCM link and IFN-α-induced downregulation of CD4 TSCM-specific genes in ATL patient cells. In conclusion, ex vivo IFN-α treatment exerts a pleiotropic effect on primary ATL cells, with a genetic IFN/STAT1/Fas axis determining apoptosis vs. proliferation and underscoring the CD4 TSCM model of ATL leukemogenesis.

HIV-1 Subtype Diversity and Prevalence of Primary Drug Resistance in a Single-Center Pediatric Cohort in Germany

OBJECTIVES

Data on drug-resistant mutations (DRMs) in HIV-1-infected therapy-naïve children are scarce. The aim of this study was to determine the HIV-1 subtype distribution and the prevalence of DRMs in therapy-naïve HIV-1-infected children who received routine care at the University Hospital Düsseldorf, Düsseldorf, Germany.

METHODS

Records of all HIV-1-infected children who received routine care between January 2005 and December 2015 were analyzed retrospectively. The collected data included demographics, clinical characteristics, CD4 cell count, viral load, HIV-1 subtype, and resistance genotype at baseline.

RESULTS

83 HIV-1-infected children received routine care during the observation period. HIV-1 subtypes were available in 61/83 patients (73.5%) and baseline HIV-1 resistance in 24 (29%). The prevalence of major DRMs was 29% (21% nucleoside reverse-transcriptase inhibitors [NRTIs], 12.5% non-NRTIs, and 4% protease inhibitors). Minor mutations in the protease gene were common (58%). Non-B subtypes were predominant (77%).

CONCLUSIONS

We report a predominance of non-subtype-B infections and a higher prevalence of DRMs compared to other pediatric cohorts from resource-rich settings. The difference in HIV-1 subtype distribution is due to the fact that a relevant proportion of pediatric patients in Germany are immigrants from high-prevalence settings in sub-Saharan Africa where non-B subtypes predominate.

Systems Immunology of Diabetes-Tuberculosis Comorbidity Reveals Signatures of Disease Complications

Comorbid diabetes mellitus (DM) increases tuberculosis (TB) risk and adverse outcomes but the pathological interactions between DM and TB remain incompletely understood. We performed an integrative analysis of whole blood gene expression and plasma analytes, comparing South Indian TB patients with and without DM to diabetic and non-diabetic controls without TB. Luminex assay of plasma cytokines and growth factors delineated a distinct biosignature in comorbid TBDM in this cohort. Transcriptional profiling revealed elements in common with published TB signatures from cohorts that excluded DM. Neutrophil count correlated with the molecular degree of perturbation, especially in TBDM patients. Body mass index and HDL cholesterol were negatively correlated with molecular degree of perturbation. Diabetic complication pathways including several pathways linked to epigenetic reprogramming were activated in TBDM above levels observed with DM alone. Our data provide a rationale for trials of host-directed therapies in TBDM, targeting neutrophilic inflammation and diabetic complication pathways to address the greater morbidity and mortality associated with this increasingly prevalent dual burden of communicable and non-communicable diseases.

Exploring resistance pathways for first-generation NS3/4A protease inhibitors boceprevir and telaprevir using Bayesian network learning

Resistance-associated variants (RAVs) have been shown to influence treatment response to direct-acting antivirals (DAAs) and first generation NS3/4A protease inhibitors (PIs) in particular. Interpretation of hepatitis C virus (HCV) genotypic drug resistance remains a challenge, especially in patients who previously failed DAA therapy and need to be retreated with a second DAA based regimen. Bayesian network (BN) learning on HCV sequence data from patients treated with DAAs could provide insight in resistance pathways against PIs for HCV subtypes 1a and 1b, in a similar way as applied before for HIV. The publicly available 'Rega-BN' tool chain was developed to study associative analyses for various pathogens. Our first analysis, comparing sequences from PI-naïve and PI-experienced patients, determined that NS3 substitutions R155K and V36M arise with PI-exposure in HCV1a infected patients, and were defined as major and minor resistance-associated variants respectively. NS3 variant 174H was newly identified as potentially related to PI resistance. In a second analysis, NS3 sequences from PI-naïve patients who cleared the virus during PI therapy and from PI-naïve patients who failed PI therapy were compared, showing that NS3 baseline variant 67S predisposes to treatment-failure and variant 72I to treatment success. This approach has the potential to better characterize the role of more RAVs, if sufficient therapy annotated sequence data becomes available in curated public databases. In addition, polymorphisms present in baseline sequences that predispose patients to therapy failure can be identified using this approach.

Automated prediction of HIV drug resistance from genotype data

Background

HIV/AIDS is a serious threat to public health. The emergence of drug resistance mutations diminishes the effectiveness of drug therapy for HIV/AIDS. Developing a computational prediction of drug resistance phenotype will enable efficient and timely selection of the best treatment regimens.

Results

A unified encoding of protein sequence and structure was used as the feature vector for predicting phenotypic resistance from genotype data. Two machine learning algorithms, Random Forest and K-nearest neighbor, were used. The prediction accuracies were examined by five-fold cross-validation on the genotype-phenotype datasets. A supervised machine learning approach for automatic prediction of drug resistance was developed to handle genotype-phenotype datasets of HIV protease (PR) and reverse transcriptase (RT). It predicts the drug resistance phenotype and its relative severity from a query sequence. The accuracy of the classification was higher than 0.973 for eight PR inhibitors and 0.986 for ten RT inhibitors, respectively. The overall cross-validated regression R²-values for the severity of drug resistance were 0.772–0.953 for 8 PR inhibitors and 0.773–0.995 for 10 RT inhibitors.

Conclusions

Machine learning using a unified encoding of sequence and protein structure as a feature vector provides an accurate prediction of drug resistance from genotype data. A practical webserver for clinicians has been implemented.

Analyses of HTLV-1 sequences suggest interaction between ORF-I mutations and HAM/TSP outcome

The region known as pX in the 3' end of the human T-cell lymphotropic virus type 1 (HTLV-1) genome contains four overlapping open reading frames (ORF) that encode regulatory proteins. HTLV-1 ORF-I produces the protein p12 and its cleavage product p8. The functions of these proteins have been linked to immune evasion and viral infectivity and persistence. It is known that the HTLV-1 infection does not necessarily imply the development of pathological processes and here we evaluated whether natural mutations in HTLV-1 ORF-I can influence the proviral load and clinical manifestation of HTLV-I-associated myelopathy/tropical spastic paraparesis (HAM/TSP). For that, we performed molecular characterization, datamining and phylogenetic analysis with HTLV-1 ORF-I sequences from 156 patients with negative or positive diagnosis for HAM/TSP. Our analyses demonstrated that some mutations may be associated with the outcome of HAM/TSP (C39R, L40F, P45L, S69G and R88K) or with proviral load (P34L and F61L). We further examined the presence of mutations in motifs of HBZ and observed that P45L mutation is located within the HBZ nuclear localization signal and was found more frequently between patients with HAM/TSP and high proviral load. These results indicate that some natural mutations are located in functional domains of ORF-I and suggests a potential association between these mutations and the proviral loads and development of HAM/TSP. Therefore it is necessary to conduct functional studies aimed at evaluating the impact of these mutations on the virus persistence and immune evasion.

HCV1b genome evolution under selective pressure of the cyclophilin inhibitor alisporivir during the DEB-025-HCV-203 phase II clinical trial

Major advances have revolutionized the HCV antiviral treatment field, with interferon-free combinations of direct-acting antivirals (DAAs) resulting into success rates of >90% for all HCV genotypes. Nevertheless, viral eradication at a global level stills remains challenging, stimulating the continued search for new affordable pan-genotypic drugs. To overcome selection of drug resistant variants, targeting host proteins can be an attractive mechanism of action. Alisporivir (Debio 025) is a potent pan-genotypic host-targeting antiviral agent, acting on cyclophilin A, which is necessary for HCV replication. The efficacy and safety of three different oral doses of alisporivir in combination with pegylated interferon-α2a given over a period of four weeks, was investigated in a randomized, double-blind and placebo-controlled phase IIa clinical trial, in 90 treatment-naïve subjects infected with chronic hepatitis C, wherefrom 58 HCV1b samples were selected for genetic sequencing purposes. Sequencing results were used to study the HCV genome for amino acid changes potentially related with selective pressure and resistance to alisporivir. By comparing baseline and on-treatment sequences, a large variation in proportion of amino acid changes was detected in all treatment arms. The NS5A variant D320E, which was previously identified during in vitro resistance selection and resulted in 3.6-fold reduced alisporivir susceptibility, emerged in two subjects in the alisporivir monotherapy arm. However, emergence of D320E appeared to be associated only with concurrent viral load rebound in one subject with 0.8log10IU/ml increase in HCV RNA. In general, for all datasets, low numbers of positions under positive selective pressure were observed, with no significant differences between naïve and treated sequences. Additionally, incomplete sequence information for some of the 22 patients and the low number of individuals per treatment arm, is limiting the power to assess the association of alisporivir or interferon treatment with the observed amino acid changes.

CRF19_cpx is an Evolutionary fit HIV-1 Variant Strongly Associated With Rapid Progression to AIDS in Cuba

Background

Clinicians reported an increasing trend of rapid progression (RP) (AIDS within 3 years of infection) in Cuba.

Methods

Recently infected patients were prospectively sampled, 52 RP at AIDS diagnosis (AIDS-RP) and 21 without AIDS in the same time frame (non-AIDS). 22 patients were sampled at AIDS diagnosis (chronic-AIDS) retrospectively assessed as > 3 years infected. Clinical, demographic, virological, epidemiological and immunological data were collected. Pol and env sequences were used for subtyping, transmission cluster analysis, and prediction of resistance, co-receptor use and evolutionary fitness. Host, immunological and viral predictors of RP were explored through data mining.

Findings

Subtyping revealed 26 subtype B strains, 6 C, 6 CRF18_cpx, 9 CRF19_cpx, 29 BG-recombinants and other subtypes/URFs. All patients infected with CRF19 belonged to the AIDS-RP group. Data mining identified CRF19, oral candidiasis and RANTES levels as the strongest predictors of AIDS-RP. CRF19 was more frequently predicted to use the CXCR4 co-receptor, had higher fitness scores in the protease region, and patients had higher viral load at diagnosis.

Interpretation

CRF19 is a recombinant of subtype D (C-part of Gag, PR, RT and nef), subtype A (N-part of Gag, Integrase, Env) and subtype G (Vif, Vpr, Vpu and C-part of Env). Since subtypes D and A have been associated with respectively faster and slower disease progression, our findings might indicate a fit PR driving high viral load, which in combination with co-infections may boost RANTES levels and thus CXCR4 use, potentially explaining the fast progression. We propose that CRF19 is evolutionary very fit and causing rapid progression to AIDS in many newly infected patients in Cuba.

•

We propose that CRF19 is evolutionary very fit, causing rapid progression to AIDS in many newly infected patients in Cuba.

•

CRF19 is a recombinant of subtype D, subtype A and subtype G, with a subtype D protease estimated to be particularly fit.

•

A fit protease with high viral load and co-infections, may boost RANTES levels and thus CXCR4 use, hence fast progression.

A new ensemble coevolution system for detecting HIV-1 protein coevolution

BACKGROUND

A key challenge in the field of HIV-1 protein evolution is the identification of coevolving amino acids at the molecular level. In the past decades, many sequence-based methods have been designed to detect position-specific coevolution within and between different proteins. However, an ensemble coevolution system that integrates different methods to improve the detection of HIV-1 protein coevolution has not been developed.

RESULTS

We integrated 27 sequence-based prediction methods published between 2004 and 2013 into an ensemble coevolution system. This system allowed combinations of different sequence-based methods for coevolution predictions. Using HIV-1 protein structures and experimental data, we evaluated the performance of individual and combined sequence-based methods in the prediction of HIV-1 intra- and inter-protein coevolution. We showed that sequence-based methods clustered according to their methodology, and a combination of four methods outperformed any of the 27 individual methods. This four-method combination estimated that HIV-1 intra-protein coevolving positions were mainly located in functional domains and physically contacted with each other in the protein tertiary structures. In the analysis of HIV-1 inter-protein coevolving positions between Gag and protease, protease drug resistance positions near the active site mostly coevolved with Gag cleavage positions (V128, S373-T375, A431, F448-P453) and Gag C-terminal positions (S489-Q500) under selective pressure of protease inhibitors.

CONCLUSIONS

This study presents a new ensemble coevolution system which detects position-specific coevolution using combinations of 27 different sequence-based methods. Our findings highlight key coevolving residues within HIV-1 structural proteins and between Gag and protease, shedding light on HIV-1 intra- and inter-protein coevolution.

Characterization of amino acids Arg, Ser and Thr at position 70 within HIV-1 reverse transcriptase

OBJECTIVES

The amino acid position 70 in HIV-1 reverse transcriptase (RT) plays an important role in nucleoside RT inhibitor (NRTI) resistance. K70R is part of the thymidine analog mutations, but also other amino acid changes have been associated with NRTI resistance, such as K70E and K70G. In this study, we investigated the in vivo selection of the HIV-1 RT mutations K70S and K70T and their in vitro effect on drug resistance and replication capacity.

METHODS

Recombinant viruses with RT mutations were generated to measure the in vitro drug susceptibility and replication capacity. Bayesian network analysis and three-dimensional modeling were performed to understand the selection and impact of the RT70 mutations.

RESULTS

K70S and K70T were found at a low frequency in RTI-experienced HIV-1 patients (0.10% and 0·20%). Baeyesian network learning identified no direct association with the in vivo exposure to any specific RTI. However, direct associations of K70S with mutations within the Q151M-complex and of K70T with K65R were observed. In vitro phenotypic testing revealed only minor effects of K70R/S/T as single mutations, associated with Q151M and within the context of the Q151M-complex.

DISCUSSION

These results suggest that the selection of K70S/T and their phenotypic impact are influenced by the presence of other mutations in RT. However, the low impact on in vitro phenotype here observed, alongside with the low in vivo prevalence, the exclusive direct association with known major RTI mutations and the unknown correlation with in vivo response, do not yet necessitate the inclusion of K70S/T in drug resistance interpretation systems.

Trends and predictors of transmitted drug resistance (TDR) and clusters with TDR in a local Belgian HIV-1 epidemic

We aimed to study epidemic trends and predictors for transmitted drug resistance (TDR) in our region, its clinical impact and its association with transmission clusters. We included 778 patients from the AIDS Reference Center in Leuven (Belgium) diagnosed from 1998 to 2012. Resistance testing was performed using population-based sequencing and TDR was estimated using the WHO-2009 surveillance list. Phylogenetic analysis was performed using maximum likelihood and Bayesian techniques. The cohort was predominantly Belgian (58.4%), men who have sex with men (MSM) (42.8%), and chronically infected (86.5%). The overall TDR prevalence was 9.6% (95% confidence interval (CI): 7.7-11.9), 6.5% (CI: 5.0-8.5) for nucleoside reverse transcriptase inhibitors (NRTI), 2.2% (CI: 1.4-3.5) for non-NRTI (NNRTI), and 2.2% (CI: 1.4-3.5) for protease inhibitors. A significant parabolic trend of NNRTI-TDR was found (p = 0.019). Factors significantly associated with TDR in univariate analysis were male gender, Belgian origin, MSM, recent infection, transmission clusters and subtype B, while multivariate and Bayesian network analysis singled out subtype B as the most predictive factor of TDR. Subtype B was related with transmission clusters with TDR that included 42.6% of the TDR patients. Thanks to resistance testing, 83% of the patients with TDR who started therapy had undetectable viral load whereas half of the patients would likely have received a suboptimal therapy without this test. In conclusion, TDR remained stable and a NNRTI up-and-down trend was observed. While the presence of clusters with TDR is worrying, we could not identify an independent, non-sequence based predictor for TDR or transmission clusters with TDR that could help with guidelines or public health measures.

Phenotypic susceptibility to antiretrovirals among clades C, F, and B/F recombinant antiretroviral-naive HIV type 1 strains

To evaluate antiretroviral phenotypic susceptibility of wild-type HIV-1 strains circulating in Brazil, samples from antiretroviral-naive individuals infected with subtypes C (n=16), F (n=9), or B/F (n=7), where reverse transcriptase is B and protease is F, were phenotyped using the Antivirogram Assay (Virco, Mechelen, Belgium). Reduced susceptibility to protease inhibitors (PIs) was observed in one C and three F isolates. None of these samples had any known PI resistance mutations. The phenotypic fold change to one PI was above the biological cut-off in three of 96 (3.1%) clade F phenotypic determinations and in one of 96 (1.0%) clade C. Phenotypic resistance to at least one nucleoside reverse transcriptase inhibitor (NRTI) was found for two B/F, four C, and three F isolates. The phenotypic fold change in susceptibility to NRTIs was above the cut-off value in nine of 111 (8.1%) clade C determinations, as compared to three of 63 (4.8%) for clade F and two of 49 (4.1%) for clade B. The phenotypic fold change to non-NRTI (NNRTI) was above the cut-off in seven of 32 (21.9%) of C isolates determinations, whereas none of the F isolates had a decrease of susceptibility. Only two of the 16 C samples had a known NNRTI resistance mutation. The NNRTI fold change was above the cut-off value in three of 14 (21.4%) phenotypic determinations of Brazilian B/F recombinants, representing clade B reverse transcriptase. NNRTI susceptibility should be better investigated in clade C and B/F recombinants.

HIV-1 fitness landscape models for indinavir treatment pressure using observed evolution in longitudinal sequence data are predictive for treatment failure

We previously modeled the in vivo evolution of human immunodeficiency virus-1 (HIV-1) under drug selective pressure from cross-sectional viral sequences. These fitness landscapes (FLs) were made by using first a Bayesian network (BN) to map epistatic substitutions, followed by scaling the fitness landscape based on an HIV evolution simulator trying to evolve the sequences from treatment naïve patients into sequences from patients failing treatment. In this study, we compared four FLs trained with different sequence populations. Epistatic interactions were learned from three different cross-sectional BNs, trained with sequence from patients experienced with indinavir (BNT), all protease inhibitors (PIs) (BNP) or all PI except indinavir (BND). Scaling the fitness landscape was done using cross-sectional data from drug naïve and indinavir experienced patients (Fcross using BNT) and using longitudinal sequences from patients failing indinavir (FlongT using BNT, FlongP using BNP, FlongD using BND). Evaluation to predict the failing sequence and therapy outcome was performed on independent sequences of patients on indinavir. Parameters included estimated fitness (LogF), the number of generations (GF) or mutations (MF) to reach the fitness threshold (average fitness when a major resistance mutation appeared), the number of generations (GR) or mutations (MR) to reach a major resistance mutation and compared to genotypic susceptibility score (GSS) from Rega and HIVdb algorithms. In pairwise FL comparisons we found significant correlation between fitness values for individual sequences, and this correlation improved after correcting for the subtype. Furthermore, FLs could predict the failing sequence under indinavir-containing combinations. At 12 and 48 weeks, all parameters from all FLs and indinavir GSS (both for Rega and HIVdb) were predictive of therapy outcome, except MR for FlongT and FlongP. The fitness landscapes have similar predictive power for treatment response under indinavir-containing regimen as standard rules-based algorithms, and additionally allow predicting genetic evolution under indinavir selective pressure.

Treatment-associated polymorphisms in protease are significantly associated with higher viral load and lower CD4 count in newly diagnosed drug-naive HIV-1 infected patients

BACKGROUND

The effect of drug resistance transmission on disease progression in the newly infected patient is not well understood. Major drug resistance mutations severely impair viral fitness in a drug free environment, and therefore are expected to revert quickly. Compensatory mutations, often already polymorphic in wild-type viruses, do not tend to revert after transmission. While compensatory mutations increase fitness during treatment, their presence may also modulate viral fitness and virulence in absence of therapy and major resistance mutations. We previously designed a modeling technique that quantifies genotypic footprints of in vivo treatment selective pressure, including both drug resistance mutations and polymorphic compensatory mutations, through the quantitative description of a fitness landscape from virus genetic sequences.

RESULTS

Genotypic correlates of viral load and CD4 cell count were evaluated in subtype B sequences from recently diagnosed treatment-naive patients enrolled in the SPREAD programme. The association of surveillance drug resistance mutations, reported compensatory mutations and fitness estimated from drug selective pressure fitness landscapes with baseline viral load and CD4 cell count was evaluated using regression techniques. Protease genotypic variability estimated to increase fitness during treatment was associated with higher viral load and lower CD4 cell counts also in treatment-naive patients, which could primarily be attributed to well-known compensatory mutations at highly polymorphic positions. By contrast, treatment-related mutations in reverse transcriptase could not explain viral load or CD4 cell count variability.

CONCLUSIONS

These results suggest that polymorphic compensatory mutations in protease, reported to be selected during treatment, may improve the replicative capacity of HIV-1 even in absence of drug selective pressure or major resistance mutations. The presence of this polymorphic variation may either reflect a history of drug selective pressure, i.e. transmission from a treated patient, or merely be a result of diversity in wild-type virus. Our findings suggest that transmitted drug resistance has the potential to contribute to faster disease progression in the newly infected host and to shape the HIV-1 epidemic at a population level.

Using iterative ridge regression to explore associations between conditioned variables

We address a specific case of joint probability mapping, where the information presented is the probabilistic associations of random variables under a certain condition variable (conditioned associations). Bayesian and dependency networks graphically map the joint probabilities of random variables, though both networks may identify associations that are independent of the condition (background associations). Since the background associations have the same topological features as conditioned associations, it is difficult to discriminate between conditioned and non-conditioned associations, which results in a major increase in the search space. We introduce a modification of the dependency network method, which produces a directed graph, containing only condition-related associations. The graph nodes represent the random variables and the graph edges represent the associations that arise under the condition variable. This method is based on ridge-regression, where one can utilize a numerically robust and computationally efficient algorithm implementation. We illustrate the method's efficiency in the context of a medically relevant process, the emergence of drug-resistant variants of human immunodeficiency virus (HIV) in drug-treated, HIV-infected people. Our mapping was used to discover associations between variants that are conditioned by the initiation of a particular drug treatment regimen. We have demonstrated that our method can recover known associations of such treatment with selected resistance mutations as well as documented associations between different mutations. Moreover, our method revealed novel associations that are statistically significant and biologically plausible.

HIV-1 mutational pathways under multidrug therapy

BACKGROUND

Genotype-derived drug resistance profiles are a valuable asset in HIV-1 therapy decisions. Therapy decisions could be further improved, both in terms of predicting length of current therapy success and in preserving followup therapy options, through better knowledge of mutational pathways- here defined as specific locations on the viral genome which, when mutant, alter the risk that additional specific mutations arise. We limit the search to locations in the reverse transcriptase region of the HIV-1 genome which host resistance mutations to nucleoside (NRTI) and non-nucleoside (NNRTI) reverse transcriptase inhibitors (as listed in the 2008 International AIDS Society report), or which were mutant at therapy start in 5% or more of the therapies studied.

METHODS

A Cox proportional hazards model was fit to each location with the hazard of a mutation at that location during therapy proportional to the presence/absence of mutations at the remaining locations at therapy start. A pathway from preexisting to occurring mutation was indicated if the covariate was both selected as important via smoothly clipped absolute deviation (a form of regularized regression) and had a small p-value. The Cox model also allowed controlling for non-genetic parameters and potential nuisance factors such as viral resistance and number of previous therapies. Results were based on 1981 therapies given to 1495 distinct patients drawn from the EuResist database.

RESULTS

The strongest influence on the hazard of developing NRTI resistance was having more than four previous therapies, not any one existing resistance mutation. Known NRTI resistance pathways were shown, and previously speculated inhibition between the thymidine analog pathways was evidenced. Evidence was found for a number of specific pathways between NRTI and NNRTI resistance sites. A number of common mutations were shown to increase the hazard of developing both NRTI and NNRTI resistance. Viral resistance to the therapy compounds did not materially effect the hazard of mutation in our model.

CONCLUSIONS

The accuracy of therapy outcome prediction tools may be increased by including the number of previous treatments, and by considering locations in the HIV genome which increase the hazard of developing resistance mutations.

Models of RNA virus evolution and their roles in vaccine design

Viruses are fast evolving pathogens that continuously adapt to the highly variable environments they live and reproduce in. Strategies devoted to inhibit virus replication and to control their spread among hosts need to cope with these extremely heterogeneous populations and with their potential to avoid medical interventions. Computational techniques such as phylogenetic methods have broadened our picture of viral evolution both in time and space, and mathematical modeling has contributed substantially to our progress in unraveling the dynamics of virus replication, fitness, and virulence. Integration of multiple computational and mathematical approaches with experimental data can help to predict the behavior of viral pathogens and to anticipate their escape dynamics. This piece of information plays a critical role in some aspects of vaccine development, such as viral strain selection for vaccinations or rational attenuation of viruses. Here we review several aspects of viral evolution that can be addressed quantitatively, and we discuss computational methods that have the potential to improve vaccine design.

Estimating the individualized HIV-1 genetic barrier to resistance using a nelfinavir fitness landscape

Background

Failure on Highly Active Anti-Retroviral Treatment is often accompanied with development of antiviral resistance to one or more drugs included in the treatment. In general, the virus is more likely to develop resistance to drugs with a lower genetic barrier. Previously, we developed a method to reverse engineer, from clinical sequence data, a fitness landscape experienced by HIV-1 under nelfinavir (NFV) treatment. By simulation of evolution over this landscape, the individualized genetic barrier to NFV resistance may be estimated for an isolate.

Results

We investigated the association of estimated genetic barrier with risk of development of NFV resistance at virological failure, in 201 patients that were predicted fully susceptible to NFV at baseline, and found that a higher estimated genetic barrier was indeed associated with lower odds for development of resistance at failure (OR 0.62 (0.45 - 0.94), per additional mutation needed, p = .02).

Conclusions

Thus, variation in individualized genetic barrier to NFV resistance may impact effective treatment options available after treatment failure. If similar results apply for other drugs, then estimated genetic barrier may be a new clinical tool for choice of treatment regimen, which allows consideration of available treatment options after virological failure.

Characterization of HIV type 1 genetic diversity among South African participants enrolled in the AIDS Vaccine Integrated Project (AVIP) study

The genetic diversity of HIV-1 strains circulating among HIV-1-infected South Africans was investigated in a cohort of 420 individuals enrolled as part of the AIDS Vaccine Integrated Project (AVIP) study. Representative samples (10%) were randomly selected from treatment-naive participants. Viral RNA was extracted for reverse transcriptase-initiated amplification and population-based sequencing of partial pol (encompassing protease and reverse transcriptase) and full-length integrase. Overall, HIV-1 sequences confirmed that 97.1% and 96.9% were HIV-1 subtype C in pol and integrase, respectively. Two participants were infected with unique A1/C and C/A1 recombinants in pol/integrase. Further pol sequence analysis identified mutation patterns associated with high level resistance to NNRTIs in two participants, whereas no primary mutations conferring resistance to integrase inhibitors were detected. The predominance of HIV-1 subtype C in South African populations is therefore confirmed in the AVIP cohort finalized for testing preventive or therapeutic vaccines against HIV-1 infection.

Resistance pathways of human immunodeficiency virus type 1 against the combination of zidovudine and lamivudine

A better understanding of human immunodeficiency virus type 1 drug-resistance evolution under the selective pressure of combination treatment is important for the design of long-term effective treatment strategies. We applied Bayesian network learning to sequences from patients treated with the reverse transcriptase inhibitor combination of zidovudine (AZT) and lamivudine (3TC) to identify the role of many treatment-selected mutations in the development of resistance. Based on the Bayesian network structure, an in vivo fitness landscape was built, reflecting the necessary selective pressure under treatment, to evolve naive sequences to sequences obtained from patients treated with the combination. This landscape, combined with an evolutionary model, was used to predict resistance evolution in longitudinal sequence pairs. In our analysis, mutations 41L, 70R, 184V and 215F/Y were identified as major resistance mutations to the combination of AZT and 3TC, as they were associated directly with treatment experience. The network also suggested a possible role in resistance development for a number of novel mutations. Estimated fitness, using the landscape, correlated significantly with in vitro resistance phenotype in genotype-phenotype pairs (R(2)=0.70). Variation in predicted evolution under selective pressure correlated significantly with observed in vivo evolution during AZT plus 3CT treatment. In conclusion, we confirmed current knowledge on resistance development to the combination of AZT and 3CT, but additional novel mutations were identified. Moreover, a model to predict resistance evolution during AZT and 3CT treatment has been built and validated.

Multi-scale modelling in computational biomedicine

The inherent complexity of biomedical systems is well recognized; they are multi-scale, multi-science systems, bridging a wide range of temporal and spatial scales. This article reviews the currently emerging field of multi-scale modelling in computational biomedicine. Many exciting multi-scale models exist or are under development. However, an underpinning multi-scale modelling methodology seems to be missing. We propose a direction that complements the classic dynamical systems approach and introduce two distinct case studies, transmission of resistance in human immunodeficiency virus spreading and in-stent restenosis in coronary artery disease.

Pairwise and higher-order correlations among drug-resistance mutations in HIV-1 subtype B protease

Background

The reaction of HIV protease to inhibitor therapy is characterized by the emergence of complex mutational patterns which confer drug resistance. The response of HIV protease to drugs often involves both primary mutations that directly inhibit the action of the drug, and a host of accessory resistance mutations that may occur far from the active site but may contribute to restoring the fitness or stability of the enzyme. Here we develop a probabilistic approach based on connected information that allows us to study residue, pair level and higher-order correlations within the same framework.

Results

We apply our methodology to a database of approximately 13,000 sequences which have been annotated by the treatment history of the patients from which the samples were obtained. We show that including pair interactions is essential for agreement with the mutational data, since neglect of these interactions results in order-of-magnitude errors in the probabilities of the simultaneous occurence of many mutations. The magnitude of these pair correlations changes dramatically between sequences obtained from patients that were or were not exposed to drugs. Higher-order effects make a contribution of as much as 10% for residues taken three at a time, but increase to more than twice that for 10 to 15-residue groups. The sequence data is insufficient to determine the higher-order effects for larger groups. We find that higher-order interactions have a significant effect on the predicted frequencies of sequences with large numbers of mutations. While relatively rare, such sequences are more prevalent after multi-drug therapy. The relative importance of these higher-order interactions increases with the number of drugs the patient had been exposed to.

Conclusion

Correlations are critical for the understanding of mutation patterns in HIV protease. Pair interactions have substantial qualitative effects, while higher-order interactions are individually smaller but may have a collective effect. Together they lead to correlations which could have an important impact on the dynamics of the evolution of cross-resistance, by allowing the virus to pass through otherwise unlikely mutational states. These findings also indicate that pairwise and possibly higher-order effects should be included in the models of protein evolution, instead of assuming that all residues mutate independently of one another.

Mutation T74S in HIV-1 subtype B and C proteases resensitizes them to ritonavir and indinavir and confers fitness advantage

OBJECTIVES

Several drug resistance and secondary mutations have been described in HIV-1 viruses from patients undergoing antiretroviral therapy. In this study, we assessed the impact of the protease substitution T74S on the phenotype and on the replicative fitness in HIV-1 subtypes B and C.

METHODS

HIV-1 molecular clones carrying subtype B or C proteases had these coding regions subjected to site-directed mutagenesis to include T74S alone or in combination with four known protease inhibitor (PI) primary drug resistance mutations. All clones were used in a phenotypic assay to evaluate their susceptibility to most commercially available PIs. The impact of T74S on virus fitness was also assessed for all viruses through head-to-head competitions and oligonucleotide ligation assays to measure the proportion of each virus in culture.

RESULTS

Viruses of both subtypes carrying T74S did not have their susceptibility altered to any tested PI. Viruses with the four resistance mutations showed strong resistance to most PIs with fold changes ranging from 5 to 300 times compared with their wild-type counterparts. Surprisingly, the addition of T74S to the multiresistant clones restored their susceptibilities to indinavir and ritonavir and partially to lopinavir, close to those of wild-type viruses. Most 74S-containing viruses were more fit than their 74T counterparts.

CONCLUSIONS

Our results suggest that T74S is not a major drug resistance mutation, but it resensitizes multiresistant viruses to certain PIs. T74S is a bona fide accessory mutation, restoring fitness of multidrug-resistant viruses in both subtypes B and C. T74S should be further studied in clinical settings and considered in drug resistance interpretation algorithms.

Effect of human immunodeficiency virus type 1 protease inhibitor therapy and subtype on development of resistance in subtypes B and G

Europe is currently observing a significant rise in non-B subtypes. Consequently, the effect of genetic variability on therapy response or genotypic resistance interpretation algorithms is an emerging concern. The purpose of this study is to investigate the amino acid substitutions selected under drug pressure in the protease of human immunodeficiency virus type 1 (HIV-1) subtypes B and G, and determine if there are any significant differences. We investigated therapy-related and subtype-related substitutions in the protease, considering subtype, overall protease inhibitor treatment and individual drug exposure. Many mutations were significantly related to protease inhibitor (PI) therapy, with mutations exclusive to subtype B or subtype G. Some mutations are at positions related to resistance in both subtypes, but the amino acid substitution is different. Other mutations were significantly associated with subtype and PI selective pressure (p<0.05), pointing towards a differential selective pressure in both subtypes. We confirmed previous reports on the subtype-dependent selection of D30N and 89I, and identified a new mutation with such differential selective pressure: 37D was preferentially selected by lopinavir in subtype B. Other novel mutations found under therapy pressure were 13A, 35N, K55R, I66F, I72L/T, T74S, 82M and 89I/V. Our study indicates that even though in general, drug selective pressure and resistance pathways are relatively similar between subtypes B and G, some differences do occur, leading to subtype-dependent substitutions.

Positive selection pressure introduces secondary mutations at Gag cleavage sites in human immunodeficiency virus type 1 harboring major protease resistance mutations

Human immunodeficiency virus type 1 (HIV-1) protease inhibitors (PIs) specifically target the HIV-1 protease enzyme. Mutations in the enzyme can result in PI resistance (termed PI mutations); however, mutations in the HIV-1 gag region, the substrate for the protease enzyme, might also lead to PI resistance. We analyzed gag and pol sequence data from the following 313 HIV-1-infected patients: 160 treatment-naïve patients, 93 patients failing antiretroviral treatment that included a PI (with no major PI mutations), and 60 patients failing antiretroviral treatment that included a PI (with major PI mutations). Additional sequences from 13 patients were included for longitudinal analysis. We assessed positive selection pressure on the gag/protease region using a test for the overall influence of positive selection and a total of five tests to identify positively selected single codons. We found that positive selection pressure was the driving evolutionary force for the gag region in all three patient groups. An increase in positive selection was observed in gag cleavage site regions p7/p1/p6 only after the acquisition of major PI mutations, suggesting that amino acids in gag cleavage sites under positive selection pressure could function as compensatory mutations for major PI mutations in the protease region. Isolated gag mutations did not appear to confer PI resistance, but mutations in the gag cleavage sites could substitute for minor PI resistance mutations in the protease region.

Novel recombinant virus assay for measuring susceptibility of human immunodeficiency virus type 1 group M subtypes to clinically approved drugs

Combination therapy can successfully suppress human immunodeficiency virus (HIV) replication in patients but selects for drug resistance, requiring subsequent resistance-guided therapeutic changes. This report describes the development and validation of a novel assay that offers a uniform method to measure susceptibility to all clinically approved HIV type 1 (HIV-1) drugs targeting reverse transcriptase (RT), protease (PR), integrase (IN), and viral entry. It is an assay in which the antiviral effect on infection within a single replication cycle is measured in triply transfected U87.CD4.CXCR4.CCR5 cells, based on homologous recombination between patient-derived amplicons and molecular proviral clones tagged with the enhanced green fluorescent protein (EGFP) reporter gene and from which certain viral genomic regions are removed. The deletions stretch from p17 codon 7 to PR codon 98 in pNL4.3-DeltagagPR-EGFP, from PR codons 1 to 99 in pNL4.3-DeltaPR-EGFP, from RT codons 1 to 560 in pNL4.3-DeltaRT-EGFP, from IN codons 1 to 288 in pNL4.3-DeltaIN-EGFP, and from gp120 codon 34 to gp41 codon 237 in pNL4.3-Deltaenv-EGFP. The optimized experimental conditions enable the investigation of patient samples regardless of viral subtype or coreceptor use. The extraction and amplification success rate for a set of clinical samples belonging to a broad range of HIV-1 group M genetic forms (A-J, CRF01-03, CRF05, and CRF12-13) and displaying a viral load range of 200 to >500,000 RNA copies/ml was 97%. The drug susceptibility measurements, based on discrimination between infected and noninfected cells on a single-cell level by flow cytometry, were reproducible, with coefficients of variation for resistance ranging from 7% to 31%, and were consistent with scientific literature in terms of magnitude and specificity.

Improved identification of conserved cassette exons using Bayesian networks

Background

Alternative splicing is a major contributor to the diversity of eukaryotic transcriptomes and proteomes. Currently, large scale detection of alternative splicing using expressed sequence tags (ESTs) or microarrays does not capture all alternative splicing events. Moreover, for many species genomic data is being produced at a far greater rate than corresponding transcript data, hence in silico methods of predicting alternative splicing have to be improved.

Results

Here, we show that the use of Bayesian networks (BNs) allows accurate prediction of evolutionary conserved exon skipping events. At a stringent false positive rate of 0.5%, our BN achieves an improved true positive rate of 61%, compared to a previously reported 50% on the same dataset using support vector machines (SVMs). Incorporating several novel discriminative features such as intronic splicing regulatory elements leads to the improvement. Features related to mRNA secondary structure increase the prediction performance, corroborating previous findings that secondary structures are important for exon recognition. Random labelling tests rule out overfitting. Cross-validation on another dataset confirms the increased performance. When using the same dataset and the same set of features, the BN matches the performance of an SVM in earlier literature. Remarkably, we could show that about half of the exons which are labelled constitutive but receive a high probability of being alternative by the BN, are in fact alternative exons according to the latest EST data. Finally, we predict exon skipping without using conservation-based features, and achieve a true positive rate of 29% at a false positive rate of 0.5%.

Conclusion

BNs can be used to achieve accurate identification of alternative exons and provide clues about possible dependencies between relevant features. The near-identical performance of the BN and SVM when using the same features shows that good classification depends more on features than on the choice of classifier. Conservation based features continue to be the most informative, and hence distinguishing alternative exons from constitutive ones without using conservation based features remains a challenging problem.

Phylogenetic dependency networks: inferring patterns of CTL escape and codon covariation in HIV-1 Gag

HIV avoids elimination by cytotoxic T-lymphocytes (CTLs) through the evolution of escape mutations. Although there is mounting evidence that these escape pathways are broadly consistent among individuals with similar human leukocyte antigen (HLA) class I alleles, previous population-based studies have been limited by the inability to simultaneously account for HIV codon covariation, linkage disequilibrium among HLA alleles, and the confounding effects of HIV phylogeny when attempting to identify HLA-associated viral evolution. We have developed a statistical model of evolution, called a phylogenetic dependency network, that accounts for these three sources of confounding and identifies the primary sources of selection pressure acting on each HIV codon. Using synthetic data, we demonstrate the utility of this approach for identifying sites of HLA-mediated selection pressure and codon evolution as well as the deleterious effects of failing to account for all three sources of confounding. We then apply our approach to a large, clinically-derived dataset of Gag p17 and p24 sequences from a multicenter cohort of 1144 HIV-infected individuals from British Columbia, Canada (predominantly HIV-1 clade B) and Durban, South Africa (predominantly HIV-1 clade C). The resulting phylogenetic dependency network is dense, containing 149 associations between HLA alleles and HIV codons and 1386 associations among HIV codons. These associations include the complete reconstruction of several recently defined escape and compensatory mutation pathways and agree with emerging data on patterns of epitope targeting. The phylogenetic dependency network adds to the growing body of literature suggesting that sites of escape, order of escape, and compensatory mutations are largely consistent even across different clades, although we also identify several differences between clades. As recent case studies have demonstrated, understanding both the complexity and the consistency of immune escape has important implications for CTL-based vaccine design. Phylogenetic dependency networks represent a major step toward systematically expanding our understanding of CTL escape to diverse populations and whole viral genes.

Drug resistance differences among HIV types and subtypes: a growing problem

Although HIV-1 subtype B accounts for only 10% of worldwide HIV infections, almost all knowledge regarding antiretroviral (ARV) drug development and viral resistance is based on this subtype. More recently, an increasing body of evidence suggests that distinct HIV genetic variants possess different biological properties, including susceptibility and response to ARVs. In this review, we will summarize recent in vitro and in vivo studies reporting such differences. In general terms, infections with most HIV variants respond well to ARVs, but minor differences in susceptibility, in the emergence and selection of subtype-specific drug resistance mutations and in the acquisition of similar mutations over the period of ARV exposure have been reported. Such differences impact on drugresistance interpretation algorithms, which are mostly based on inference from sequence information. Despite the differences observed, clinical response to ARV therapy among subjects infected with distinct HIV variants is effective, and the dissemination of ARV access in developing countries where non-B subtypes prevail should not be delayed.

Bayesian network analyses of resistance pathways against efavirenz and nevirapine

OBJECTIVE

To clarify the role of novel mutations selected by treatment with efavirenz or nevirapine, and investigate the influence of HIV-1 subtype on nonnucleoside reverse transcriptase inhibitor (nNRTI) resistance pathways.

DESIGN

By finding direct dependencies between treatment-selected mutations, the involvement of these mutations as minor or major resistance mutations against efavirenz, nevirapine, or coadministrated nucleoside analogue reverse transcriptase inhibitors (NRTIs) is hypothesized. In addition, direct dependencies were investigated between treatment-selected mutations and polymorphisms, some of which are linked with subtype, and between NRTI and nNRTI resistance pathways.

METHODS

Sequences from a large collaborative database of various subtypes were jointly analyzed to detect mutations selected by treatment. Using Bayesian network learning, direct dependencies were investigated between treatment-selected mutations, NRTI and nNRTI treatment history, and known NRTI resistance mutations.

RESULTS

Several novel minor resistance mutations were found: 28K and 196R (for resistance against efavirenz), 101H and 138Q (nevirapine), and 31L (lamivudine). Robust interactions between NRTI mutations (65R, 74V, 75I/M, and 184V) and nNRTI resistance mutations (100I, 181C, 190E and 230L) may affect resistance development to particular treatment combinations. For example, an interaction between 65R and 181C predicts that the nevirapine and tenofovir and lamivudine/emtricitabine combination should be more prone to failure than efavirenz and tenofovir and lamivudine/emtricitabine.

CONCLUSION

Bayesian networks were helpful in untangling the selection of mutations by NRTI versus nNRTI treatment, and in discovering interactions between resistance mutations within and between these two classes of inhibitors.

A genotypic assay for the amplification and sequencing of integrase from diverse HIV-1 group M subtypes

Recently, the Food and Drug Administration (FDA) of the USA approved the first integrase inhibitor for inclusion in treatment regimens of HIV-1 patients failing their current regimens with multi-drug resistant strains. However, treatment failure has been observed during integrase inhibitor-containing therapy. Several mutational pathways have been described with signature mutations at integrase positions 66, 92, 148 and 155. Therefore, a genotypic assay for the amplification and sequencing of HIV-1 integrase was developed. The assay displayed a detection limit of 10 HIV-1 III(B) RNA copies/ml plasma. As the HIV-1 pandemic is characterised by a large genetic diversity, the new assay was evaluated on a panel of 74 genetically divergent samples belonging to the following genetic forms A, B, C, D, F, G, J, CRF01-AE, CRF02-AG, CRFF03-AB, CRF12-BF and CRF13-cpx. Their viral load ranged from 178 until >500,000 RNA copies/ml. The amplification and sequencing was successful for 70 samples (a success rate of 95%). The four failures were most probably due to low viral load or poor quality of RNA and not to subtype issues. Some of the sequences obtained from integrase inhibitor-naïve patients displayed polymorphisms at integrase positions associated with resistance: 74IV, 138D, 151I, 157Q and 163AE. The relevance of these polymorphisms in the absence of the signature mutations remains unclear.

Modelled in vivo HIV Fitness under drug Selective Pressure and Estimated Genetic Barrier Towards Resistance are Predictive for Virological Response

Background

A method has been developed to estimate a fitness landscape experienced by HIV-1 under treatment selective pressure as a function of the genotypic sequence thereby also estimating the genetic barrier to resistance.

Methods

We evaluated the performance of two estimated fitness landscapes (nelfinavir [NFV] and zidovudine [AZT] plus lamivudine [3TC]) to predict week 12 viral load (VL) change for 176 treatment change episodes (TCEs) and probability of week 48 virological failure for 90 TCEs, in treatment experienced patients starting these drugs in combination.

Results

A higher genetic barrier for AZT plus 3TC, (quantified per additional mutation required to develop resistance against these drugs) was associated with a 0.54 (95% confidence interval [CI] 0.30–0.77) larger log10 VL reduction at 12 weeks ( P<0.0001) and a 0.39 (95% CI 0.23–0.66) lower odds of virological failure at 48 weeks ( P=0.0005), in analyses adjusting for the pre-TCE VL and the exact time-lag between the TCE and the date of determining response VL. The strength of these associations was comparable with those seen with expert interpretation systems (Rega, ANRS and HIVDB). A higher genetic barrier to NFV resistance was the only genotypic predictor that tended to be associated with a 0.19 (95% CI 0–0.39) higher log10 VL reduction at 12 weeks ( P=0.05) and a 0.63 (95% CI 0.36–1.09) lower odds of virological failure at 48 weeks ( P=0.10) per additional mutation.

Conclusions

These results suggest that an estimated genetic barrier derived from fitness landscapes may contribute to an improvement of predicted treatment outcome for NFV and this approach should be explored for other drugs.

HIV-1 gp120 N-linked glycosylation differs between plasma and leukocyte compartments

Background

N-linked glycosylation is a major mechanism for minimizing virus neutralizing antibody response and is present on the Human Immunodeficiency Virus (HIV) envelope glycoprotein. Although it is known that glycosylation changes can dramatically influence virus recognition by the host antibody, the actual contribution of compartmental differences in N-linked glycosylation patterns remains unclear.

Methodology and Principal Findings

We amplified the env gp120 C2-V5 region and analyzed 305 clones derived from plasma and other compartments from 15 HIV-1 patients. Bioinformatics and Bayesian network analyses were used to examine N-linked glycosylation differences between compartments. We found evidence for cellspecific single amino acid changes particular to monocytes, and significant variation was found in the total number of N-linked glycosylation sites between patients. Further, significant differences in the number of glycosylation sites were observed between plasma and cellular compartments. Bayesian network analyses showed an interdependency between N-linked glycosylation sites found in our study, which may have immense functional relevance.

Conclusion

Our analyses have identified single cell/compartment-specific amino acid changes and differences in N-linked glycosylation patterns between plasma and diverse blood leukocytes. Bayesian network analyses showed associations inferring alternative glycosylation pathways. We believe that these studies will provide crucial insights into the host immune response and its ability in controlling HIV replication in vivo. These findings could also have relevance in shielding and evasion of HIV-1 from neutralizing antibodies.

Estimation of an in vivo fitness landscape experienced by HIV-1 under drug selective pressure useful for prediction of drug resistance evolution during treatment

MOTIVATION

HIV-1 antiviral resistance is a major cause of antiviral treatment failure. The in vivo fitness landscape experienced by the virus in presence of treatment could in principle be used to determine both the susceptibility of the virus to the treatment and the genetic barrier to resistance. We propose a method to estimate this fitness landscape from cross-sectional clinical genetic sequence data of different subtypes, by reverse engineering the required selective pressure for HIV-1 sequences obtained from treatment naive patients, to evolve towards sequences obtained from treated patients. The method was evaluated for recovering 10 random fictive selective pressures in simulation experiments, and for modeling the selective pressure under treatment with the protease inhibitor nelfinavir.

RESULTS

The estimated fitness function under nelfinavir treatment considered fitness contributions of 114 mutations at 48 sites. Estimated fitness correlated significantly with the in vitro resistance phenotype in 519 matched genotype-phenotype pairs (R(2) = 0.47 (0.41 - 0.54)) and variation in predicted evolution under nelfinavir selective pressure correlated significantly with observed in vivo evolution during nelfinavir treatment for 39 mutations (with FDR = 0.05).

AVAILABILITY

The software is available on request from the authors, and data sets are available from http://jose.med.kuleuven.be/~kdforc0/nfv-fitness-data/.

An evolutionary-network model reveals stratified interactions in the V3 loop of the HIV-1 envelope

The third variable loop (V3) of the human immunodeficiency virus type 1 (HIV-1) envelope is a principal determinant of antibody neutralization and progression to AIDS. Although it is undoubtedly an important target for vaccine research, extensive genetic variation in V3 remains an obstacle to the development of an effective vaccine. Comparative methods that exploit the abundance of sequence data can detect interactions between residues of rapidly evolving proteins such as the HIV-1 envelope, revealing biological constraints on their variability. However, previous studies have relied implicitly on two biologically unrealistic assumptions: (1) that founder effects in the evolutionary history of the sequences can be ignored, and; (2) that statistical associations between residues occur exclusively in pairs. We show that comparative methods that neglect the evolutionary history of extant sequences are susceptible to a high rate of false positives (20%-40%). Therefore, we propose a new method to detect interactions that relaxes both of these assumptions. First, we reconstruct the evolutionary history of extant sequences by maximum likelihood, shifting focus from extant sequence variation to the underlying substitution events. Second, we analyze the joint distribution of substitution events among positions in the sequence as a Bayesian graphical model, in which each branch in the phylogeny is a unit of observation. We perform extensive validation of our models using both simulations and a control case of known interactions in HIV-1 protease, and apply this method to detect interactions within V3 from a sample of 1,154 HIV-1 envelope sequences. Our method greatly reduces the number of false positives due to founder effects, while capturing several higher-order interactions among V3 residues. By mapping these interactions to a structural model of the V3 loop, we find that the loop is stratified into distinct evolutionary clusters. We extend our model to detect interactions between the V3 and C4 domains of the HIV-1 envelope, and account for the uncertainty in mapping substitutions to the tree with a parametric bootstrap.

Mapping protease inhibitor resistance to human immunodeficiency virus type 1 sequence polymorphisms within patients

Resistance genotyping provides an important resource for the clinical management of patients infected with human immunodeficiency virus type 1 (HIV-1). However, resistance to protease (PR) inhibitors (PIs) is a complex phenotype shaped by interactions among nearly half of the residues in HIV-1 PR. Previous studies of the genetic basis of PI resistance focused on fixed substitutions among populations of HIV-1, i.e., host-specific adaptations. Consequently, they are susceptible to a high false discovery rate due to founder effects. Here, we employ sequencing "mixtures" (i.e., ambiguous base calls) as a site-specific marker of genetic variation within patients that is independent of the phylogeny. We demonstrate that the transient response to selection by PIs is manifested as an excess of nonsynonymous mixtures. Using a sample of 5,651 PR sequences isolated from both PI-naive and -treated patients, we analyze the joint distribution of mixtures and eight PIs as a Bayesian network, which distinguishes residue-residue interactions from direct associations with PIs. We find that selection for resistance is associated with the emergence of nonsynonymous mixtures in two distinct groups of codon sites clustered along the substrate cleft and distal regions of PR, respectively. Within-patient evolution at several positions is independent of PIs, including those formerly postulated to be involved in resistance. These positions are under strong positive selection in the PI-naive patient population, implying that other factors can produce spurious associations with resistance, e.g., mutational escape from the immune response.

Mining complex genotypic features for predicting HIV-1 drug resistance

MOTIVATION

Human immunodeficiency virus type 1 (HIV-1) evolves in human body, and its exposure to a drug often causes mutations that enhance the resistance against the drug. To design an effective pharmacotherapy for an individual patient, it is important to accurately predict the drug resistance based on genotype data. Notably, the resistance is not just the simple sum of the effects of all mutations. Structural biological studies suggest that the association of mutations is crucial: even if mutations A or B alone do not affect the resistance, a significant change might happen when the two mutations occur together. Linear regression methods cannot take the associations into account, while decision tree methods can reveal only limited associations. Kernel methods and neural networks implicitly use all possible associations for prediction, but cannot select salient associations explicitly.

RESULTS

Our method, itemset boosting, performs linear regression in the complete space of power sets of mutations. It implements a forward feature selection procedure where, in each iteration, one mutation combination is found by an efficient branch-and-bound search. This method uses all possible combinations, and salient associations are explicitly shown. In experiments, our method worked particularly well for predicting the resistance of nucleotide reverse transcriptase inhibitors (NRTIs). Furthermore, it successfully recovered many mutation associations known in biological literature.

AVAILABILITY

http://www.kyb.mpg.de/bs/people/hiroto/iboost/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

HIV-1 subtype B protease and reverse transcriptase amino acid covariation

Despite the high degree of HIV-1 protease and reverse transcriptase (RT) mutation in the setting of antiretroviral therapy, the spectrum of possible virus variants appears to be limited by patterns of amino acid covariation. We analyzed patterns of amino acid covariation in protease and RT sequences from more than 7,000 persons infected with HIV-1 subtype B viruses obtained from the Stanford HIV Drug Resistance Database (http://hivdb.stanford.edu). In addition, we examined the relationship between conditional probabilities associated with a pair of mutations and the order in which those mutations developed in viruses for which longitudinal sequence data were available. Patterns of RT covariation were dominated by the distinct clustering of Type I and Type II thymidine analog mutations and the Q151M-associated mutations. Patterns of protease covariation were dominated by the clustering of nelfinavir-associated mutations (D30N and N88D), two main groups of protease inhibitor (PI)-resistance mutations associated either with V82A or L90M, and a tight cluster of mutations associated with decreased susceptibility to amprenavir and the most recently approved PI darunavir. Different patterns of covariation were frequently observed for different mutations at the same position including the RT mutations T69D versus T69N, L74V versus L74I, V75I versus V75M, T215F versus T215Y, and K219Q/E versus K219N/R, and the protease mutations M46I versus M46L, I54V versus I54M/L, and N88D versus N88S. Sequence data from persons with correlated mutations in whom earlier sequences were available confirmed that the conditional probabilities associated with correlated mutation pairs could be used to predict the order in which the mutations were likely to have developed. Whereas accessory nucleoside RT inhibitor-resistance mutations nearly always follow primary nucleoside RT inhibitor-resistance mutations, accessory PI-resistance mutations often preceded primary PI-resistance mutations.

BioBayesNet: a web server for feature extraction and Bayesian network modeling of biological sequence data

BioBayesNet is a new web application that allows the easy modeling and classification of biological data using Bayesian networks. To learn Bayesian networks the user can either upload a set of annotated FASTA sequences or a set of pre-computed feature vectors. In case of FASTA sequences, the server is able to generate a wide range of sequence and structural features from the sequences. These features are used to learn Bayesian networks. An automatic feature selection procedure assists in selecting discriminative features, providing an (locally) optimal set of features. The output includes several quality measures of the overall network and individual features as well as a graphical representation of the network structure, which allows to explore dependencies between features. Finally, the learned Bayesian network or another uploaded network can be used to classify new data. BioBayesNet facilitates the use of Bayesian networks in biological sequences analysis and is flexible to support modeling and classification applications in various scientific fields. The BioBayesNet server is available at http://biwww3.informatik.uni-freiburg.de:8080/BioBayesNet/.

Impact of genetic variation of HIV-1 on drug resistance development

HIV is a particularly diverse virus and can be classified into different types, groups and subtypes. Most of the current knowledge is based on HIV-1 subtype B, which is responsible for the first recognized epidemic, mainly spreading in the developed world, while the other subtypes are responsible for the major part of the pandemic, primarily in the developing world. Interest in the so-called ‘non-B’ subtypes is increasing, with treatment programs expanding to include Africa and Asia. According to several recent studies, differences exist between subtypes regarding drug susceptibility, resistance development and therapy response. Currently, conclusions are often based on too few patients or samples, and the use of different resistance assays and analysis methods makes it difficult to compare results. A common mistake is to pool subtypes and recombinants together as ‘non-B subtypes’, although they differ among each other as much as they differ from subtype B. Europe, with large epidemics of subtype B and several non-B subtypes in similar patient populations, is well placed to take the lead in this research field.