点击显示 收起
Divisions of Communicable Disease and Immunology and Retrovirology, Walter Reed Army Institute of Research, Washington, DC
Henry M. Jackson Foundation for the Advancement of Military Medicine, Rockville, Maryland
Background.
An assessment of biomarkers from an analysis of human peripheral blood mononuclear cell gene-expression profiles was made, to acquire an understanding of transcriptional changes associated with human immunodeficiency virus type 1 (HIV-1) infection in vivo.
Methods.
Supervised learning algorithms were used to create signature gene sets that could be used to distinguish seropositive from seronegative samples and delineate changes in disease status during the early stages of infection. Bioinformatic tools were used to classify persons and to functionally characterize groups of differentially expressed genes, to elucidate the impact of viral infection on host cell gene-expression patterns.
Results.
A 10-gene signature set that could be used to accurately determine the HIV-1 serostatus was identified. A 6-gene signature set was used to distinguish seropositive persons exhibiting differential changes in CD4+ T cell counts, with 93% accuracy. Functional classification of differentially expressed genes in HIV-1 indicated a preponderance of down-regulated genes with functions related to the immune response and apoptosis. Hierarchical cluster analysis in persons whose CD4+ T cell counts increased, compared with that in persons whose CD4+ T cell counts decreased, was characterized by the down-regulation of genes associated with apoptosis, mitochondrial function, protein biosynthesis, and RNA binding.
Conclusions.
Gene-expression profile analysis of a complex infectious virus, such as HIV-1, may be useful to elucidate the functional genomic relationships associated with viral infection.
The wide range of clinical manifestations of HIV-1 infection is the result of a complex set of host-virus interactions [1, 2]. Classically, clinical markers used to assess the state of infection involve static assessment of CD4+ T cell counts and plasma viral loads. Particularly useful to efforts to elucidate the mechanisms of HIV-1 viral pathogenesis is the study of groups of seropositive persons exhibiting divergent rates of disease progression during the 510 years after seroconversion [36]. Results of such studies indicate that long-term survival involves maintenance of a low viral load by a strong virus-specific immune response [79]. Importantly, rates of disease progression can be modified, at least temporarily, by the diligent use of antiretroviral drugs to lengthen survival and reduce morbidity [10, 11]. Lacking, however, are biomarkers that might be used to prognosticate the rate of HIV-1 disease progression and, by extension, the success of drug or vaccine intervention administered earlier during disease and treatment.
Precedent for the use of cell-associated biomarkers to prognosticate disease progression and response to treatment exists in the implementation of gene-expression profile analysis to classify certain cancers and tumor types [1218]. Extending the successful use, in oncology, of cellular expressionbased molecular classification systems to an infectious disease, we used gene-expression profile analysis to examine primary peripheral blood mononuclear cells (PBMCs) from HIV-1seropositive and seronegative persons, to (1) determine the fundamental gene-expression signature that can be used to classify a sample according to its serostatus, (2) classify samples as being from persons with divergent degrees of change in disease status, and (3) examine functional classes of genes impacted by HIV-1 infection in the context of what is known about the disease.
SUBJECTS, MATERIALS, AND METHODS
Clinical specimens.
PBMCs were obtained from seropositive persons who provided informed consent and who were enrolled in studies approved by local institutional review boards. Confirmed-seronegative samples were clinical discards from the mandatory US Military Force Screening Program that evaluates the serostatus of active personnel on a regular basis. Peripheral blood was collected, by venipuncture, in acid citrate dextrose, and PBMCs were separated by Ficoll-gradient (Sigma) centrifugation and were cryopreserved. Plasma aliquots were stored at -80°C.
Preparation of samples for GeneChip analysis.
Preparation of cellular RNA for GeneChip analysis, cDNA preparation and in vitro transcription, and staining and scanning of Affymetrix Human Focus GeneChips (Affymetrix) were performed essentially as described elsewhere [19].
Plasma viral load assessment.
The Roche AMPLICOR HIV Monitor test (version 1.5; Roche Diagnostics) was used to quantify the amount of HIV-1 RNA in plasma. All persons had a viral load above the cutoff value (>50 copies/mL) of the ultrasensitive test.
Repository information.
The Affymetrix data sets can be accessed at http://www.ncbi.nlm.nih.gov/geo/, under the accession number GSE2171.
GeneChip quality control.
Two criteria were used to determine GeneChip quality: (1) the scaling factor, determined by use of the Affymetrix Microarray Suite 5.0, with target signal intensity set to its default (500), and (2) the array outlier percentage, determined by use of dChip (version 1.3) [20]. CEL files were normalized at the probe level by use of the robust multichip average method [21]. Genes (probe sets) that had >50% "absent" calls were filtered out.
Classification and prediction.
Twenty-two seropositive and 7 seronegative samples were used as a training set to generate a set of genes that could be used to determine the serostatus of an unclassified sample, as described by Ramaswamy et al. [22]. To determine the optimal number of genes to be included in a predictor set, a weighted-voting classification algorithm was applied. The leave-one-out cross-validation method was used to simulate classification accuracy for the top 100 genes on the basis of a signal-to-noise statistic. Simulation results indicated the top 10 genes that achieved 100% classification accuracy in a training set and that were subsequently evaluated in a test set. These 10 genes composed the serostatus predictor set. The same approach was used to generate a 6-gene signature set to be used to distinguish samples from seropositive persons with differential changes in CD4+ T cell counts during the study period.
Differential regulation.
Differentially regulated genes were identified by use of the statistical program Significance Analysis of Microarrays (SAM) [23], and cluster analysis of microarray data sets was performed by use of Cluster and Treeview software (available at: http://rana.lbl.gov/EisenSoftware.htm).
Derivation of gene ontology and functional associations.
The functions and biological classifications of differentially regulated gene sets were further analyzed by use of the Web-based tools Onto-Express (available at: http://vortex.cs.wayne.edu:8080/index.jsp) [24] and Gene Ontology Tree Machine (available at: http://genereg.ornl.gov/gotm/) [25]. PathwayAssist (version 2.5; Ariadne Genomics) software was used for analysis of the biological pathway.
RESULTS
Characteristics of the study groups.
A total of 87 primary clinical samples consisting of human PBMCs were used in the present study, including 12 seronegative samples from healthy control subjects, 22 seropositive samples from drug-naive persons, 21 seropositive samples from persons who had received at least 1 antiretroviral drug regimen, and 32 seropositive samples from persons whose CD4+ T cell counts either decreased or increased during the study period. Seropositive persons with differential changes in CD4+ T cell counts may have received nucleoside reverse-transcriptase inhibitors (NRTIs) but not highly active antiretroviral therapy (HAART). These study groups were defined ad hoc, and samples were drawn from a specimen repository.
Table 1 shows descriptive statistics for the 22 seropositive samples from drug-naive persons. This group is a cross-sectional cohort of seropositive persons distinguished by having high, medium, or low plasma viral loads.
Table 2 shows descriptive statistics for the seropositive persons with differential changes in CD4+ T cell counts. In the present study, these 2 groups of persons with changes in CD4+ T cell counts were defined by the magnitude and direction of the change in CD4+ T cell counts. Emphasis was placed on the transcriptional profiling of the early stages of infection in persons whose CD4+ T cell counts either increased or decreased between 2 consecutive time points during a 27-month period. There was no clinically relevant difference between the 2 groups during the interval between seroconversion and time point 1 (TP1), either for those persons whose CD4+ T cell counts decreased (mean ± SE, 1.0 ± .07 years) or for those persons whose CD4+ T cell counts increased (mean ± SE, 1.3 ± .03 years). There was no statistically significant difference between the 2 groups in terms of CD4+ T cell counts at TP1. Persons whose CD4+ T cell counts decreased were defined by a decrease of 8.29 CD4+ T cells/mL/month, whereas persons whose CD4+ T cell counts increased were defined by an increase of 4.80 CD4+ T cells/mL/month. Plasma viral loads at time point 2 (TP2) were significantly different between the 2 groups (mean ± SE, 89,075 ± 13,920 copies/mL in the group with decreasing CD4+ T cell counts vs. 10,461 ± 1961 copies/mL in the group with increasing CD4+ T cell counts; P = .01). The antiretroviral drug treatment regimens for persons in both groups are presented in table 3.
By use of supervised learning algorithms and a leave-one-out cross-validation method, the 10-gene signature set derived from the training set was applied to an independent test set of samples that were blinded with regard to HIV-1 serostatus, to assess the accuracy of classification. The results of the classification analysis for both the training set and the test set are shown as 3-dimensional principal-component plots (figure 1C). An accuracy of 93% was demonstrated for the 58 samples included in the test set. Four of the samples that were classified as seronegative were misclassified and, when unblinded, were determined to be seropositive. Although the training set was composed entirely of samples from persons who were completely drug naive, the test set was composed of samples from persons who could have been receiving any antiretroviral treatment. As this set was the most clinically relevant set, we felt that this conservative analysis was the most appropriate method to classify HIV-1positive samples.
Longitudinal analysis revealed striking differences in gene expression between the 2 groups. The majority of differentially expressed genes that changed over time were found exclusively in samples from persons whose CD4+ T cell counts increased (420 genes), compared with those whose CD4+ T cell counts decreased (15 genes). Gene-expression values for the 420 genes in the longitudinal data set and the 531 genes in the cross-sectional data set were expressed as log2 fold changes relative to the mean expression values from 12 seronegative persons and are shown as the average gene-expression ratio, in a centroid plot (figure 3D). The changes in expression that distinguished the 2 groups from each other were characterized by a switch in the pattern of gene expression from one of similar early induction in both groups to transcriptional repression in those persons whose CD4+ T cell counts increased.
(1 of 2 images)
DISCUSSION
We determined whether a set of biomarkers that separate persons on the basis of their HIV-1 serostatus, independent of the amount of viremia present, could be identified. We found that host transcriptional profiling by use of unsupervised and supervised clustering methodologies could be used to classify persons according to serostatus, with a high degree of accuracy. The observations made in the present study are derived from an assessment of the PBMC compartment, which represents a highly heterogeneous mixture of cell types. Although sampling from this compartment facilitates access to clinical specimens, it may not reflect gene-expression patterns that are associated with specific cell types or processes confined to tissue-associated germinal centers known to be important in HIV-1 disease. Furthermore, gene-expression patterns that we identified by use of the Affymetrix GeneChip platform, a highly redundant oligomer array system, may differ from those identified by use of cDNA arrays or other array platforms [26].
Analysis of samples from drug-naive persons showed that >97% of all differentially expressed genes were down-regulated or underexpressed, compared with those from seronegative persons. Assessment of the functions of genes whose expression is impacted by HIV-1 revealed clusters of genes involved in cell proliferation and the nucleosome. Changes in chromatin structure and nucleosome remodeling include processes that are linked to transcriptional regulation by altering DNA replication and gene expression through accessibility to transcription factors, activators, and repressors involved both in host cell activity and regulation of HIV-1 gene expression [27, 28].
Immune-response genes that are usually transcriptionally repressed in HIV-1seropositive patients were overrepresented in the present study. Several of the genes that were significantly repressed have previously been shown to play central immunomodulatory roles in HIV-1 infection. Examples include HLA-DRB3, HLA-DRA, HLA-DMB, HLA-DQA1, and HLA-DOB genes, in which impaired class II expression contributes to the global immunosuppression observed in HIV-1 infection [29, 30]. CD14, which was down-regulated in persons in the present study, has previously been shown to be involved in lipopolysaccharide-induced stimulation of HIV-1 replication [31]. Interleukin (IL)15 and IL-16 were likewise repressed during HIV-1 infection. Studies in nonhuman primates have demonstrated that transcription of these cytokines in macaques infected with pathogenic isolates of simian immunodeficiency virus plays a role in containing viral replication [32, 33]. IL-16 has been shown to repress HIV infection in lymphocytes and monocytes by inhibiting viral transcription.
Viruses have evolved strategies to evade immune responses by enhancing viral replication and decreasing host cell survival, by modulation of genes of the NF- transcription-factor pathway [34]. A set of genes involved in the positive regulation of the I- kinase/NF- cascade was found to be repressed during HIV infection. NF- activation is induced by I- kinases, which are involved in orchestrating the host immune response against infection while, at the same time, promoting HIV-1 replication. In drug-naive seropositive persons, modulation of the NF- pathways during infection can also profoundly affect the pathways that influence the host cell cycle and regulation of cellular proliferation.
Transcriptional repression of signal transductionpathway genes involved in the regulation of NF- by HIV-1 can have dual, opposing effects, such as accelerating apoptosis or protecting cells from programmed cell death. Thus, it is not surprising to find that, in the present study, transcriptional activation or repression of functional clusters of differentially expressed genes involved in programmed cell death were identified as prominent features of HIV-1 infection. Our observations extend many of those made in gene-expression profile analyses of lymph node samples from seropositive persons receiving aggressive HAART regimens, in that productive viral infection is associated with the modulation of genes associated with the immune function, activation, and apoptosis [35].
We next explored the potential of using host biomarkers derived from gene-expression profile analysis to distinguish persons with differential changes in CD4+ T cell counts during the early stages of infection. PBMCs from these persons were successfully classified by use of a set of 6 predictor genes. It is necessary to consider that, although the selection of a small number of genes may yield a high degree of accuracy when used for classification purposes, such a limited set of genes is probably not reflective of the complex gene functions and interactions that distinguish persons with different disease states.
Analysis of gene-expression profiles in these 2 groups revealed several features that were not observed in investigation of differentially expressed genes from the drug-naive group. We found that the persons whose CD4+ T cell counts increased exhibited the most change in gene expression during the study period. By contrast, between TP1 and TP2, very few genes from persons whose CD4+ T cell counts decreased showed changes in expression. Differential gene expression between the 2 groups was greater at TP2 than at TP1, logically reflecting the study design, which closely matched persons at TP1. By TP2, the persons in the 2 progression groups differed significantly with respect to CD4+ T cell counts, plasma viral loads, and associated cellular gene-expression profiles. This observation was surprising, since it exposed a seemingly contradictory divergence in gene expression that occurred in persons whose viral infection progressed or remained inactive during >2 years of surveillance. We had speculated that groups of functionally expressed genes involved in programmed cell death would be incrementally activated or induced in persons whose CD4+ T cell counts decreased over time, compared with those in persons whose CD4+ T cell counts increased. Conversely, we discovered that persons whose CD4+ T cell counts increased had the most-acute changes in gene expression. These changes were reflected in clusters of statistically significant down-regulated genes belonging to functionally overrepresented groups involved in processes related to cell death. These genes are associated with mitochondrial functions, including electron transport, conversion of NADH to ubiquinone, and cytochrome c oxidase activity. The observations that genes associated with the mitochondria were differentially expressed must be qualified in the context that NRTIs are known to induce mitochondrial toxicity [36]. However, since the use of NRTIs in the 2 study groups was equivalent, the identification of differentially expressed mitochondrial genes associated with apoptosis remains significant.
The finding that biological processes, such as programmed cell death, distinguished persons with differential changes in CD4+ T cell counts was further supported by construction of a gene-interaction network by use of PathwayAssist. A gene that formed a central node in this biological-interaction network was TNFRSF6 (CD95 or FADD) [37]. That productive infection with HIV-1 resulted in down-regulation of the immune response and modulation of apoptosis may reflect mechanisms in the host cell that either foster viral production in a stable, otherwise healthy cellular host or control infection by down-regulating genes required for disease progression.
The present study has demonstrated that a complex infectious virus, such as HIV-1, can be effectively characterized by the use of gene-expression profile analysis and is a significant step toward the derivation of a set of biomarkers based on differential levels of transcription that may become powerful tools in the assessment of interventions that foster a slower rate of disease progression, especially when used with classic markers, such as CD4+ T cell count and viral load. Finally, identification of specific gene pathways impacted by viral infection may lead to the use of less-toxic targeted therapies. Further evaluation of the prognostic power of gene-expression analysis in HIV-1 disease awaits studies of greater numbers of persons.
Acknowledgments
We thank Dr. Deborah L. Birx (Director of the Military HIV-1 Research Program), for support of this effort, and Drs. Nelson Michael and Mark Lewis, for helpful discussions. Expert research of clinical data records and technical laboratory work in support of this study were executed by Eva Calero (Walter Reed Army Institute of Research) and Martin Nau, Alma Arnold, and Caroline Liebig (Henry M. Jackson Foundation).
References
1. Hogan CM, Hammer SM. Host determinants of HIV-1 infection and disease. Part 1: cellular and humoral immune responses. Ann Intern Med 2001; 134:76176. First citation in article
2. Hogan CM, Hammer SM. Host determinants of HIV infection and disease. Part 2: genetic factors and implications for antiretroviral therapeutics. Ann Intern Med 2001; 134:97896. First citation in article
3. Hogervorst E, Jurriaans S, deWolf F, et al. Predictors for non- and slow progression in HIV-1 infection: low viral copy numbers in serum and maintenance of high HIV-1 p24-specific but not V3-specific antibody levels. J Infect Dis 1995; 171:81121. First citation in article
4. Giorgi JV, Lylees RH, Matud JL, et al. Predictive value of immunologic and virologic markers after long or short duration of HIV-1 infection. J Acquir Immune Defic Syndr 2002; 29:34655. First citation in article
5. Rubio CM, Rubio RC, Nogues BA, Falguerra SM, Manonelles FA. Survival and disease progression in 251 patients with HIV-1 infection: a study of p24 antigen and viral burden as prognosis markers and their value at 4 years of follow-up. An Med Interna 2001; 18:51720. First citation in article
6. Kim S, Hughes MD, Hammer SM, Jackson JB, DeGruttola V, Katzenstein DA. Both serum HIV type 1 RNA levels and CD4+ lymphocyte counts predict clinical outcome in HIV type 1 infected subjects with 200 to 500 CD4+ cell per cubic millimeter. AIDS Clinical Trials Group Study 175 Virology Study Team. AIDS Res Hum Retroviruses 2000; 16:64553. First citation in article
7. Cao Y, Qin L, Zhang L, Safrit J, Ho DH. Virologic and immunologic characterization of long term survivors of human immunodeficiency virus type 1 infection. N Engl J Med 1995; 332:2018. First citation in article
8. Pantaleo G, Menzo S, Vaccarrezza M, et al. Studies in subjects with long term nonprogressive human immunodeficiency virus infection. N Engl J Med 1995; 332:20916. First citation in article
9 Cecilia D, Kleeberger C, Munoz A, Giorgo JV, Zoller-Panzer S. A longitudinal study of neutralizing antibodies and disease progression in HIV-1infected subjects. J Infect Dis 1999; 179:136574. First citation in article
10. Hogg RS, Yip B, Chan KJ, et al. Rates of disease progression by baseline CD4 cell count and viral load after initiating triple drug therapy. JAMA 2001; 286:256877. First citation in article
11. Chene G, Sterne JA, May M, et al. Prognostic importance of initial response in HIV-1 infected patients starting potent antiretroviral therapy: analysis of prospective studies. Lancet 2003; 362:67986. First citation in article
12. Huang E, Cheng SH, Dressman H, et al. Gene expression predictors of breast cancer outcomes. Lancet 2003; 361:15906. First citation in article
13. Van de Vijver MJ, He YD, Van't Veer LJ, et al. A gene expression signature as a predictor of survival in breast cancer. N Engl J Med 2002; 347:19992009. First citation in article
14. Hofmann WK, deVos S, Elashoff D, et al. Relation between resistance of Philadelphia chromosome positive acute lymphoblastic leukemia to the tyrosine kinase inhibitor STI571 and gene expression profiles: a gene expression study. Lancet 2002; 359:4816. First citation in article
15. Iizuka N, Masaaki O, Yamada-Okabe H, et al. Oligonucleotide microarray for prediction of early intrahepatic recurrence of hepatocellular carcinoma after curative resection. Lancet 2003; 361:9239. First citation in article
16. Bullinger L, Dohner K, Bair E, et al. Use of gene expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N Engl J Med 2004; 350:160516. First citation in article
17. Valk PJM, Verhaak RGW, Beijen MA, et al. Prognostically useful gene expression profiles in acute myeloid leukemia. N Engl J Med 2004; 350:161728. First citation in article
18. Yeoh EJ, Ross ME, Shurtleff SA, et al. Classification, subtype discovery, and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling. Cancer Cell 2002; 1:13343. First citation in article
19. Vahey MT, Nau ME, Taubman M, Yalley-Ogunro J, Silvera P, Lewis MG. Patterns of gene expression in peripheral blood mononuclear cells of rhesus macaques infected with SIVmac251 and exhibiting differential rates of disease progression. AIDS Res Hum Retroviruses 2003; 19:36987. First citation in article
20. Li C, Wong W. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc Natl Acad Sci USA 2001; 98:316. First citation in article
21. Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, Speed TP. Summaries of Affymetrix GeneChip probe level data. Nucl Acids Res 2003; 31:e15. First citation in article
22. Ramaswamy S, Ross KN, Lander ES, Golub TR. A molecular signature of metastasis in primary solid tumors. Nat Genet 2003; 33:4954. First citation in article
23. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci USA 2001; 98:511621. First citation in article
24. Draghici S, Khatri P, Bhavsar P, Shah A, Krawetz S, Tainsky MA. Onto-Tools, the toolkit of the modern biologist: Onto-Express, Onto-Compare, Onto-Design and Onto-Translate. Nucleic Acids Res 2003; 31:377581. First citation in article
25. Zhang B, Schmoyer D, Kirov S, Snoddy J. GOTree Machine (GOTM): a Web-based platform for interpreting sets of interesting genes using Gene Ontology hierarchies. BMC Bioinformatics 2004; 5:16. First citation in article
26. Tan PK, Downey TJ, Spitznagel EL, et al. Evaluation of gene expression measurements from commercial microarray platforms. Nucleic Acid Res 2003; 31:567684. First citation in article
27. Ito T. Nuclesome assembly and remodeling. Curr Top Microbiol Immunol 2003; 274:122. First citation in article
28. He G, Ylisastigui L, Margolis DM. The regulation of HIV-1 gene expression: the emerging role of chromatin. DNA Cell Biol 2002; 21:697705. First citation in article
29. Schindler M, Wurfl S, Benaroch P. Down-modulation of mature major histocompatibility complex class II and up-regulation of invariant chain cell surface expression are well-conserved functions of human and simian immunodeficiency virus nef alleles. J Virol 2003; 77:1054856. First citation in article
30. Polyak S, Chen H, Hirsch D, George I, Hershberg R, Sperber K. Impaired class II expression and antigen uptake in monocytic cells after HIV-1 infection. J Immunol 1997; 159:217788. First citation in article
31. Bagasra O, Wright SD, Seshamma T, Oakes JW, Pomerantz RJ. CD14 is involved in control of human immunodeficiency virus type 1 expression in latently infected cells by lipopolysaccharide. Proc Natl Acad Sci USA 1992; 89:62859. First citation in article
32. Caufour P, Le Grand R, Cheret A, et al. Longitudinal analysis of CD8+ T-cell phenotype and IL-7, IL-15 and IL-16 mRNA expression in different tissues during primary simian immunodeficiency virus infection. Microbes Infect 2001; 3:18191. First citation in article
33. Ahmad R, Sindhu ST, Toma E, Morisset R, Ahmad A. Studies on the production of IL-15 in HIV-infected/AIDS patients. J Clin Immunol 2003; 23:8190. First citation in article
34. Hiscott J, Kwon H, Genin P. Hostile takeovers: viral appropriation of the NF-kappaB pathway. J Clin Invest 2001; 107:14351. First citation in article
35. Li Q, Schacker T, Carlis J, Beilman G, Nguyen P, Haase AT. Functional genomic analysis of the response of HIV-1infected lymphatic tissue to antiretroviral therapy. J Infect Dis 2004; 189:57282. First citation in article
36. Lewis W, Dalakas MC. Mitochondrial toxicity of antiretroviral drugs. Nat Med 1995; 1:41722. First citation in article
37. Genini D, Sheeter D, Rought S, et al. HIV induces lymphocyte apoptosis by a p53-initiated, mitochondrial-mediated mechanism. FASEB J 2001; 15:56. First citation in article