Articles Epigenetic landscape in the kick-and-kill therapeutic vaccine BCN02 clinical trial is associated with antiretroviral treatment interruption (ATI) outcome Bruna Oriol-Tordera,a,b Anna Esteve-Codina,c,d María Berdasco,e,f Míriam Rosa s-Umbert,a,g Elena Gonçalves,h a,b  a a Clara Duran-Castells, Francesc Catala-Moll, Anuska Llano, Samandhy Ceden ~o,a Maria C. Puertas,a,i Martin Tolstrup,g as Hanke,m,n Behazine Combadiere,h Roger Paredes,a,i,j,k Ole S. Søgaard,g Bonaventura Clotet,a,i,j,k Javier Martínez-Picado,a,i,k,l Tom Dennis Hartigan-O’Connor,o,p Manel Esteller,l,q,r,s Michael Meulbroek,t María Luz Calle,u Alex Sanchez-Pla,v,w Jose Molto ,i,j a,i,j,k a,k,l a,i Beatriz Mothe, Christian Brander, and Marta Ruiz-Riol * a IrsiCaixa, AIDS Research Institute, Institute for Health Science Research Germans Trias i Pujol (IGTP), Hospital Germans Trias i Pujol, 2nd floor, Ctra del canyet s/n, Badalona, Barcelona 08916, Spain b Departament de Biologia Cel¢lular, de Fisiologia i d'Immunologia, Universitat Auto noma de Barcelona, Av. de Can Domenech 737, Cerdanyola del Vall es, Barcelona 08193, Spain c Centro Nacional de Analisis Geno mico (CNAG), Barcelona Science Park - Tower I, Carrer de Baldiri Reixac 4, Barcelona 08028, Spain d Universitat Pompeu Fabra (UPF), Plaça de la Merc e 10-12, Barcelona 08002, Spain e Cancer Epigenetics and Biology Program (PEBC), Bellvitge Biomedical Research Institute, Vinguda de la Granvia de l’Hospi- talet 199, L'Hospitalet de Llobregat, Barcelona 08907, Spain f Epigenetic Therapies Group, Experimental and Clinical Hematology Program (PHEC), Josep Carreras Leukaemia Research Institute, Ctra de Can Ruti - Camí de les Escoles, s/n, Badalona, Barcelona 08916, Spain g Department of Clinical Medicine - Department of Infectious Disease, Aarhus University Hospital, Palle Juul-Jensens Boule- vard 99, East Jutland, Aarhus 8200, Denmark h Centre d’Immunologie et des Maladies Infectieuses Paris (Cimi-Paris), INSERM U1135, Sorbonne Universit e, Bd de l'H^ opital 91, Paris, ^Ile de France 75013, France i CIBERINFEC, Madrid, Spain j Fundacio  Lluita contra la Sida, Infectious Diseases Department, Hospital Universitari Germans Trias i Pujol, Ctra del Canyet s/ n, Badalona, Barcelona 08916, Spain k Centre for Health and Social Care Research (CESS), Faculty of Medicine, University of Vic - Central University of Catalonia (UVic - UCC), Carrer Miquel Martí i Pol, 1, Vic, Barcelona 08500, Spain l Institucio Catalana de Recerca i Estudis Avançats (ICREA), Passeig de Lluís Companys, 23, Barcelona 08010, Spain m The Jenner Institute, University of Oxford, Old Road Campus Research Build, Roosevelt Dr, Headington, Oxford, Oxfordshire OX3 7DQ, UK n Joint Research Center for Human Retrovirus Infection, Kumamoto University, Honjo 2-2-1, Kumamoto City, Chuo-ku 860- 0811, Japan o Department of Medical Microbiology and Immunology, University of California, Veterinary Medicine 3A, Davis, CA 95616, USA p Division of Experimental Medicine, UC Davis School of Medicine, 4610 X Street, Sacramento, CA 95817, USA q Cancer and Leukemia Epigenetics and Biology Program (PEBCL), Josep Carreras Leukaemia Research Institute, Ctra de Can Ruti - Camí de les Escoles, s/n, Badalona, Barcelona 08916, Spain r Centro de Investigacion Biomedica en Red Cancer (CIBERONC), Av. Monforte de Lemos 3-5. Pabello n 11. Planta 0, Madrid 28029, Spain s Department of Physiological Sciences II, School of Medicine, University of Barcelona, Feixa Llarga, s/n, L’Hospitalet de Llo- bregat, Barcelona 08907, Spain t Projecte dels NOMS-Hispanosida, BCN Checkpoint, Carrer del Comte Borrell, 164-166, Barcelona 08015, Spain u Biosciences Department, Faculty of Sciences and Technology, University of Vic-Central University of Catalonia, Carrer de la Laura 13 Torre dels Frares, Vic, Barcelona 08500, Spain v Statistics Department, Biology Faculty, University of Barcelona, Diagonal 643, Barcelona 08028, Spain w Statistics and Bioinformatics Unit Vall d'Hebron Institut de Recerca (VHIR), Passeig de la Vall d'Hebron, 129, Barcelona 08035, Spain *Corresponding author at: IrsiCaixa, AIDS Research Institute, Institute for Health Science Research Germans Trias i Pujol (IGTP), Hospital Germans Trias i Pujol, 2nd floor, Ctra del canyet s/n, Badalona, Barcelona 08916, Spain. E-mail address: mruiz@irsicaixa.es (M. Ruiz-Riol). www.thelancet.com Vol 78 Month April, 2022 1 Articles Summary Background The BCN02-trial combined therapeutic vaccination with a viral latency reversing agent (romidepsin, eBioMedicine 2022;78: RMD) in HIV-1-infected individuals and included a monitored antiretroviral pause (MAP) as an efficacy read-out 103956 identifying individuals with an early or late (< or > 4weeks) viral-rebound. Integrated -omics analyses were applied Published online xxx prior treatment interruption to identify markers of virus control during MAP. https://doi.org/10.1016/j. ebiom.2022.103956 Methods PBMC, whole-genome DNA methylation and transcriptomics were assessed in 14 BCN02 participants, including 8 Early and 4 Late viral-rebound individuals. Chromatin state, histone marks and integration analysis (his- tone-3 acetylation (H3Ac), viral load, proviral levels and HIV-specific T cells responses) were included. REDUC-trial samples (n = 5) were included as a control group for RMD administration alone. Findings DNA methylation imprints after receiving the complete intervention discriminated Early versus Late viral- rebound individuals before MAP. Also, differential chromatin accessibility and histone marks at DNA methylation level were detected. Importantly, the differential DNA methylation positions (DMPs) between Early and Late rebounders before MAP were strongly associated with viral load, proviral levels as well as the HIV-specific T-cell responses. Most of these DMPs were already present prior to the intervention and accentuated after RMD infusion. Interpretation This study identifies host DNA methylation profiles and epigenetic cascades that are predictive of subsequent virus control in a kick-and-kill HIV cure strategy. Funding European Union Horizon 2020 Framework Programme for Research and Innovation under Grant Agree- ment N°681137-EAVI2020 and N°847943-MISTRAL, the Ministerio de Ciencia e Innovaci on (SAF2017_89726_R), and the National Institutes of Health National Institute of Allergy and Infectious Diseases Program Grant P01- AI131568. Copyright Ó 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/) Keywords: HIV-1 vaccine; Epigenetics; DNA methylation pathways related to HIV-1 infection (“HIV life cycle”), Research in context immune response (“T-cell response”, “B cell response” and “innate immunity”) and other relevant biological Evidence before this study processes like cell proliferation, gene expression regula- tion and metabolism. Additionally, we identify that the The lack of an effective HIV-1 cure is mainly due to the epigenomic landscape after the whole intervention and virus capacity to establish HIV-1 reservoirs within before treatment interruption is associated with the dif- latently infected cells. To overcome this issue, the kick- ferential viral rebound kinetics and is additionally corre- and-kill strategy has been proposed as a way to disrupt lated with HIV-1 proviral levels, peripheral blood viral the HIV-1 latency and to improve the antiviral immune loads and the strength of the HIV-1-specific T-cell response. The BCN02 study, combined a therapeutic responses, suggesting an important role of host epige- HIV-1 vaccine with romidepsin, a latency reversing nomic regulation to achieve an HIV-1 cure. agent, and showed that some individuals maintain a low viral load for extended periods of time after antire- Implications of all the available evidence troviral treatment interruption (ATI). The understanding of the mechanisms and host factors driving such HIV-1 Our findings, together with a recent article in indi- control could benefit new HIV-1 cure strategies. Since viduals treated with a latency reversing agent previous studies of gene expression profiles and DNA (NCT01680094), which identified DNA methylation methylation have yielded promising imprints associated imprints associated with HIV-1 reservoir size and with natural HIV-1 control and potential therapeutic tar- time to viral rebound after ATI, suggest that the gets, a combined transcriptomics and epigenomics inclusion of DNA methylation studies in future clini- study was conducted in the context of the BCN02 study. cal studies might be relevant to identify predictors of viral control prior to the interruption of antiretro- Added value of this study viral treatment. Additionally, a few studies have related HIV-1 infection, ART treatment, and HIV-1 In the present study, we determined that the adminis- control with DNA methylation marks, therefore a tration of an HIV-1 therapeutic vaccine together with comprehensive study of the host epigenome in clini- the latency reversing agent romidepsin has a profound cal intervention might be relevant to fine-tune such impact on the host epigenomics and transcriptomics strategies to induce effective HIV cure. patterns, which ultimately affect genes involved in 2 www.thelancet.com Vol 78 Month April, 2022 Articles Introduction the entire 32 weeks of MAP without the need to restart Human immunodeficiency virus-1 (HIV-1) infection ART.13 To gain insight into the molecular processes forms a viral reservoir during early stages of infection modulated during the clinical study, especially those which makes lifelong antiretroviral treatment (ART) that can affect HIV-1-specific T-cell responses and virus indispensable and presents the main barrier to an HIV- reactivation, as well as to identify predictors of ATI out- 1 cure. Additionally, despite the effectiveness of ART, it come before ART interruption, a systems biology analy- is not available worldwide, it is costly and is associated sis was conducted combining whole-genome gene with viral resistance and treatment-related side effects. expression and DNA methylation assessments in The establishment of an HIV-1 reservoir along with PBMCs samples from 14 individuals enrolled in the sub-optimal and dysfunctional antiviral immune BCN02 clinical trial at three different time points responses lays the basis of uncontrolled viral replication (Figure S1a). and HIV disease progression in the absence of ART. To Recent findings of specific DNA methylation pat- this end, different kick-and-kill strategies are being terns associated with the natural ability to control HIV designed to disrupt viral latency and enhance the hosts replication in vivo,16,17 raised the questions as to immune system to eliminate HIV-1-infected cells.1 3 whether stable methylation profiles prior to ART inter- A wide variety of latency reversing agents (LRAs) ruption could determine the treatment outcome, and have been proposed to reactivate the viral reservoir, whether kick-and-kill strategies would have the power including epigenetic modifiers and agonists of NF-kB, to restore potentially epigenetically dysregulated host Toll-Like-Receptor (TLR) and protein kinase C, among gene pathways important for virus control. Therefore, others.4 Histone deacetylase inhibitors (HDACi) like DNA methylation was evaluated together with gene vorinostat, panabinostat or romidepsin (RMD), origi- expression data in the context of the BCN02 trial. These nally developed for other medical indications (cancer), results revealed the main pathways that were modulated were first to be translated to the HIV cure field.4,5 by the cumulative effect of the vaccination and RMD, as HDACi administration is associated with a global well as their impact on chromatin states and histone increase of histone acetylation that leads to chromatin marks. Despite the small number of participants, spe- relaxation and an increase in gene expression, including cific DNA methylation imprints in pre-MAP samples HIV transcripts, thus rendering virally infected cells vis- that discriminate Early (n = 8) or Late (n = 4) rebound- ible to the immune system. However, LRA administra- ing individuals were identified. Such marks were related tion has generally failed to eliminate HIV-infected cells to relevant pathways for virus infection (HIV infection, efficiently and no clinically relevant reductions in provi- HIV life cycle and Host interactions of HIV infection), and rus loads have been observed.4,6,7 Given that HDACi to immune signatures (TCR signalling, MHC-II antigen were developed as anticancer drugs, they may hinder presentation and signalling by interleukins); and mapped the effective clearance of reactivated virus-infected cells to regions with differential chromatin relaxation states by a transient immune-suppression of CD8 T cell activ- and transcription factor binding sites (TFBS). Future ity, whose effects would be cumulative with a generally analyses in larger clinical trials, ideally double-blinded, dysfunctional CTL immune response due to viral and placebo-controlled, will be needed to confirm these infection.8 10 To overcome these limitations, clinical tri- results and to test the use of DNA methylation signa- als combining LRAs with therapeutic T cell vaccination tures as a predictive tool for viral rebound before treat- aim to control rebounding HIV by intensifying the anti- ment interruptions are initiated. HIV specific CTL responses.11 13 However, the effects of LRAs administration on non-viral, host genes involved in antiviral immunity are poorly understood, Methods although they may play an important role in shaping effective vaccine-induced immunity. Study design BCN02 (NCT02616874) was a pilot study that com- BCN02 (NCT02616874) was a pilot kick-and-kill study bined the MVA.HIVconsv14 vaccination before and after that included 15 early-treated individuals who were the administration of three cycles of RMD in early rolled-over from the prior clinical trial BCN01 treated individuals rolled-over from the BCN01 (NCT01712425) (Table S1).15 In BCN01, individuals (NCT01712425) clinical trial.13,15 In BCN02, RMD led to were immunized with a heterologous prime/boost regi- a transient increase of HIV transcription and histone 3 men delivering the HIVconsv immunogen. In BCN02, acetylation levels (H3Ac) that were short-lived and the participants were vaccinated again with MVA.HIV- returned to basal levels one week after the last RMD consv before and after the administration of RMD (1 cycle.13 The BCN02 study also included a 32-week moni- week after 3 cycles of RMD infusion) (Figure S1a). The tored antiretroviral pause (MAP) to evaluate virus con- BCN02 included a monitored antiretroviral pause trol post-intervention. Of the 13 individuals eligible for (MAP) to evaluate the viral rebound kinetics whereby MAP, 4 maintained viral loads < 2000 HIV RNA cop- participants were monitored for viral load rebounds on ies/mL for more than 4 weeks and of these, 3 completed a weekly basis for up to 32 weeks. Two participants were www.thelancet.com Vol 78 Month April, 2022 3 Articles excluded from MAP evaluations, one who did not meet lane, following the manufacturer’s protocol. Image anal- the criteria for treatment interruption, and one who ysis, base calling and quality scoring of the run were showed a protocol deviation during this phase. Specifi- processed using the manufacturer’s software Real Time cally, the present omics sub-study was based on the 14 Analysis (RTA 1.18.66.3) and followed by generation of male individuals enrolled in the BCN02, of which 12 FASTQ sequence files by CASAVA. RNA-seq paired- were eligible for MAP. These 12 individuals were end reads were mapped against the human (GRCh38) classed in Early rebound (n = 8) or Late rebound group genome using STAR version 2.5.3a18 with ENCODE (n = 4) according if they reached the threshold of 2000 parameters. Annotated genes (genecode version 28) HIV RNA copies/mL and restart of ART before or after were quantified using RSEM version 1.3.0 with default the 4 weeks of treatment interruption. parameters.19 The quality control of the mapping and In order to evaluate the contribution of RMD-only quantification steps was performed with GEMtools treatment on gene expression profiles in isolated CD4 T (https://gemtools.github.io/) and custom python cells and whole PBMCs, samples from the part A of the scripts. Following RSEM expected counts were normal- clinical trial REDUC (NCT02092116)6 were used. ized with TMM method from edgeR R/Bioconductor REDUC was designed to evaluate the in vivo effect of and genes with less than 5 counts were filtered out.20 RMD administration alone (1 week after 3 cycles of Finally, counts were transformed with the voom func- RMD infusions) in ART-treated individuals and showed tion in limma R/Bioconductor and used in downstream that RMD could disrupt latency and led to a 2.4 to 5 fold analyses.21 increased HIV-1 transcription (cell-associated un-spliced HIV-1 RNA). Here, we included 5 participants from this clinical trial, sampled at 6 weeks before (Baseline) and 1 DNA methylation array week after RMD administration, consistent to the same Genomic DNA was bisulfite converted using EZ DNA timepoints used in BCN02 study (Table S2). The sam- methylation Kit (Zymo) following the manufacturer’s ple size was determined based on sample availability of protocol. Next, 4uL of bisulfite-converted DNA were the BCN02 and the REDUC studies, respectively. hybridized to Infinium HumanMethylation450 Bead- Chip following Illumina Infinium HD Methylation pro- tocol. For samples obtained during MAP, Infinium PBMCs and CD4 T cell isolation and acid nucleic MethylationEPIC BeadChip kit was used. Chip analysis extraction was performed using Illumina HiScan SQ fluorescent Blood samples from BCN02 participants were proc- scanner and the intensities of the images were extracted essed using Lymphoprep (STEMCELL technologies) using GenomeStudio (2010.3) Methylation module and 2 £ 106 dry pelleted PBMCs were frozen until use. (1.8.5) software. Quality control, background correction DNA and RNA were isolated simultaneously from the and quantile normalization across arrays was performed same sample (AllPrep DNA/RNA Mini Kit, Qiagen) using ChAMP R/Bioconductor package with functions and were frozen until use. PBMC’s were isolated from extracted from Minfi R/Bioconductor package.22 24 blood samples from REDUC trial participants using Ficoll separation followed by CD4 T-cells isolation using a CD4 T-cell isolation kit and magnetic-activated cell Differential gene expression and DNA methylation sorting (MACS) columns (Miltenyi Biotec, purity analysis >95%). Isolated CD4 cells were lysed using AllPrep In exploratory data analysis of both -omics data sets, lysis buffer and lysates were stored at -80°C until RNA MDS (Multidimensional scaling) plots were used to was extracted (AllPrep isolation kit, Qiagen). identify batch effects and the major source of variation. In the DNA methylation dataset, the only female partici- pant appeared as an outlier in the MDS plot. Addition- RNAseq library preparation and sequencing and data ally, the impossibility to take sex variation into account pre-processing due to the presence of a single female in the study, this The total RNA was quantified by QubitÒ RNA BR Assay participant was removed from the omics substudy kit (Thermo Fisher Scientific) and the RNA integrity (Figure S1b). BCN02 was conducted across two hospi- was estimated by using RNA 6000 Nano Bioanalyzer tals in the Barcelona metropolitan area and biological 2100 Assay (Agilent). The RNASeq libraries from total samples were processed at the respective laboratories. PBMC RNA samples were prepared using a TruSeqTM The sample processing laboratory parameter was also Stranded Total RNA kit protocol (Illumina) and the identified as a source of variation in gene expression rRNA was depleted from 20 to 500 ng of total RNA data (Figure S1c). Therefore, ComBat function from using the RiboZero Magnetic Gold Kit. The libraries SVA R/Biconductor was used to correct for this effect in were sequenced on HiSeq2000 (Illumina) in paired- unsupervised analysis.25 Differential gene expression end mode with a read length of 2 £ 76 bp using TruSeq and methylation analyses between time points and Early SBS Kit v4 in a fraction of a sequencing v4 flow cell and Late rebounders were performed with limma.21 For 4 www.thelancet.com Vol 78 Month April, 2022 Articles longitudinal comparisons, ‘duplicateCorr’ function was and PubMed bibliographic research and considering used to consider the participants variable as a random the context of the study and the sample type (PBMCs). effect. To identify the changes in gene expression over time, the model »Laboratory+Week was fitted, to adjust for the confounding variable Laboratory. For DNA meth- Chromatin state and histone marks enrichment ylation (M-values), the model »Week was used. Next, to To identify the chromatin regions and histone marks in identify the individuals with an Early (n = 8) or Late which DEGs or DMPs were enriched, either based on (n = 4) rebound in the different time points, the fitted differences between time points or between early or late model was »Week_Rebound. rebound, a genomic regions enrichment was run with LOLA R/Bioconductor30 based on the publicly available Roadmap datasets of 18-chromatin states Gene expression and DNA methylation profiles (ChromHMM) and histone marks of PBMCs (E062, Soft clustering (MFuzz R/Bioconductor) allowed dis- https://egg2.wustl.edu/roadmap/web_portal/chr_state_ cerning different longitudinal clusters of gene expres- learning.html). As background, non-differentially sion or DNA methylation. Briefly, genes or CpG expressed genes or the non-differentially methylated positions that showed a p-value < 0.05 in any of the CpGs positions were used. pairwise comparison between time points were included. Next, for each gene in each of the time points, Transcription factor (TF) enrichment the median expression/methylation standardized value To determine the enrichment of the TF binding site was used for clustering. The function MFuzz:mesti- motifs in DMPs between Early and Late rebound (at the mate() was used to estimate the fuzzier parameters, and Vacc+RMD time point), HOMER motif discovery soft- the number of clusters was determined with the ware was used,31 considering a 250 bp window MFuzz::Dmin() function. Finally, a minimal member- upstream and downstream of the DMPs. The other ship of 0.1 was used for gene expression, and of 0.2, for CpGs in the 450K array were used as background. DNA methylation. Evaluation of DNA methylation and gene expression Integrated functional pathways analysis of genes and capacity to discriminate between early and late CpG positions modulated during the kick-and-kill rebound therapy To test the capacity of DNA methylation and gene For integration of the differentially expressed genes expression to differentiate Early or Late rebound at the (DEGs, p-value < 0.05) and differentially methylated different time points analysed, a Principal Component positions (DMP p-value < 0.05) data sets corresponding Analysis (PCA) in each of the datasets for each time to one week after RMD administration (Vacc+RMD) point was used, and a logistic regression was fitted with and the baseline (BSL) (Figure S1), a published protocol PC1 and PC2 as predictors. Area Under de Curve (AUC) was followed.26 For functional analyses, a pre-ranked values showed how well the PC1 and PC2 classified the GSEA was run in GSEA JAVA desktop program Early or Late rebounder groups. Based on the DMPs (MacOSX Version 4.03). Genes were ranked according between Early and Late Rebound at Vacc+RMD time to the log10 p-value multiplied by the sign of log2 point, a GSEA analysis was applied using canonical Fold-Change.26 As database, a combination of the pathways and BTMs as described in methods above. REACTOME subset of Canonical pathways from Molec- Based on GSEA results, 6 modules of DMPs were iden- ular Signature database (MSigDB, v7.4), and the blood tified, and the Z-score methylation levels (M-values) of transcriptional modules (BTMs) previously described DMPs on genes in these pathways were represented in elsewhere27 were used. The majority of parameters was a heatmap (ComplexHeatmap R/Bioconductor).32 Addi- set to default (1000 permutations, weighted enrichment tionally, the Rho values (Spearman Correlation) of statistic and only gene sets between 15 and 500 genes DMPs with specific viral (HIV viral load and proviral were considered). Combined GSEA results for gene levels) parameters, HIV-1-specific T-cell responses expression and DNA methylation were plotted using (measured by INFg ELISpot) and Histone H3 acetyla- EnrichmentMap27 in Cytoscape (MacOSX version 3.7.3). tion levels were plotted. The different nodes represent the pathways, and the edges show the mutual overlap. The largest pathway was maintained when there were one or more pathways Determination of viral parameters with similar names, one embedded inside the other. Peripheral blood proviral levels (CA-HIV-1 DNA) were BTMs annotations in the miscellaneous groups TBC measured by ddPCR in isolated CD4 T cells. Proviral were as well removed.28 Finally, groups of pathways DNA and cell-associated HIV-1 RNA (CA-HIV-1 RNA) were done using AutoAnnotate29 and modified accord- were both assessed using two different sets of primers/ ing to authors criteria based on REACTOME data base probe annealing to 5’LTR and Gag. The RPP30 www.thelancet.com Vol 78 Month April, 2022 5 Articles housekeeping gene was used for genomic DNA normal- Statistics ization while TBP housekeeping gene before RMD The moderated t-test in limma was used to identify the administration was used for gene expression normaliza- differentially methylated positions or the differentially tion. To determine the HIV-1 RNA plasma levels below expressed genes between different time points or 20 copies/ml ultrasensitive viral load was performed between Early and Late rebound. Although false discov- with 4 8 ml plasma samples from participants, ultra- ery rate (FDR) adjusted p-value was calculated for each centrifuged (170,000xg, 4 °C, 30 min) prior to RNA gene transcript or CpG position, due to the small sam- viral extraction (m2000sp Abbot device). Abbott Real- ple size and the exploratory character of the present Time HIV-1 assay was used to determine HIV-1 RNA study, an uncorrected p-value < 0.05 was used to define copies. the subset of differentially expressed genes (DEGs) and differentially methylated positions (DMPs). Chi- square tests were applied to evaluate the different abundance of DMPs or DEGs in certain chromo- IFNg ELISpot assay somes in contrast to the evaluated CpG positions or T-cell responses against HIV-1 or HIVconsv were transcripts. The same test was applied for DMPs in assessed by IFNg ELISpot as described in the previous relation to island (island, open sea, shore or self) or BCN02 main study report.13 Briefly, thawed cryopre- in relation to gene (5’UTR, TSS200, TSS1500, served PBMC from samples drawn at different time 1stExon, Body, 3’UTR) and for DEGs in relation to points were cultured in duplicate in 96-well ELISpot biotypes. For chi-square test, a p-value < 0.05 was plates (MultiScreen HTS MSIPS4W10, Millipore) and considered significant. Finally, Spearman’s rank cor- stimulated with pools of 15-mer peptides overlapping relation test was used in correlation analysis and a p- peptides covering the HIVconsv immunogen (HIV- value < 0.05 was considered significant. consv, 6 pools of 32-33 peptides per pool) or the regions of HIV-1 not covered by the immunogen (12 pools of 39 47 peptides) as described.13 Medium without pepti- Ethics des was used as negative control, and as positive control The study of all samples used in the present study was the cells were stimulated with PHA and the commer- approved by the Clinical Research Ethics Committee of cially available CEF peptide pool (C.T.L OH, USA). The the Germans Trias i Pujol University Hospital (Refer- threshold to consider a positive T-cell response was the ence number PI-18-183) and all the participants signed maximum value of 3 determinations: 50 SFC (Spots an informed consent. Forming Cells)/106 PBMC (5 spots per well), > the mean of SFC in negative control wells plus 3 standard deviation of the negative control wells, or > 3 £ the Data deposition and material sharing mean of negative control wells. The HIVtotal T-cell Transcriptomics and DNA Methylation is uploaded at magnitude was determined as the sum of SFC/106 Gene Expression Omnibus (GEO): GSE184653, con- PBMC for all positive responses, the HIVtotal breadth tains the transcriptomics data for BCN02 study; was calculated as the sum of all the positive responses. GSE185391, the DNA Methylation data for BCN02 The same was applied to the responses targeting the study; and GSE185027, the transcriptomics data for the regions covered by the HIVconsv immunogen to deter- individuals from the REDUC study. The R code used in mine the magnitude and breadth of the HIVconsv spe- the analysis can be found in the GitHub https://github. cific T-cell response. com/hostimmuneOMICS under the repository BCN02_OmicsAnalysis. Histone H3 acetylation (H3Ac) Cryopreserved PBMCs were used to evaluate the H3Ac Role of the funding source in lymphocytes by flow cytometry as described previ- This publication has received funding from the Euro- ously13. Briefly, after a 20 min blocking with PBS 10% pean Union Horizon 2020 Framework Programme for FBS, cells were stained for 30 min with anti-acetyl his- Research and Innovation under Grant Agreement N° tone H3 polyclonal rabbit (MerckMillipore 06 599) or 681137-EAVI2020 and N° 847943-MISTRAL, the Min- with the control stain using normal rabbit serum (Life- isterio de Ciencia e Innovaci on (SAF2017_89726_R), Technologies 10510). After washing, a 30 min incuba- and the National Institutes of Health National Insti- tion with the secondary antibody in the dark was tute of Allergy and Infectious Diseases Program Grant performed (donkey anti-rabbit IgG, LifeTechnologies P01-AI131568. The funding sources had no role in writ- A21206). The median fluorescence intensity (MFI) for ing the manuscript nor in data collection, analysis and acetyl histone H3 stain was determined and the back- interpretation or any other aspect of the study. Funding ground was subtracted using the MFI of the isotype con- sources only require to submit the publication in open trol staining. access. 6 www.thelancet.com Vol 78 Month April, 2022 Articles Results protein-coupled receptors) as well as interleukin and cytokine signalling (Figure 1d). Correlation analysis of Gene expression and DNA methylation changes one DEGs belonging to these categories were associated week after MVA.HIVconsv vaccination in the BCN02 with the magnitude of the IFNg T-cell responses against study total HIV-1 (including the sum of HIVregions in plus To evaluate the impact of the first MVA.HIVconsv vacci- out of HIVconsv), while fewer were correlated with nation (Vacc) on the host transcriptional and epigenetic IFNg T-cell responses against HIVconsv (Figure 1e). Of signatures, whole genome gene expression and DNA these, MCRS is a gene involved in chromatin organiza- methylation screenings were conducted. Multidimen- tion, and the genes SAT1, RNFT2 and ACOT9 are all sional scaling analysis (MDS) and volcano plot of data involved in metabolic processes.33,34 from samples drawn one week after the HIV.consv vac- DNA methylation pathway analyses showed an over- cination, revealed a strong impact on the host gene tran- all negative enrichment score, with the majority of the scriptional program (Figures S2a and 1a). Specifically, gene-annotated DMPs being hypomethylated after vac- compared to basal time point, 673 upregulated cination. These DMPs were mainly involved in infec- (Figure 1a, red) and 524 downregulated (Figure 1a, blue) tion-related pathways (e.g. HIV life cycle, HCMV early genes were identified (p-value < 0.05, limma model, events) (Figure 1f). In parallel, pathways like interleukin 4 Table S3). After adjusting for multiple comparisons and 3 signalling and SLITs and ROBOs signalling con- with false discovery rate (FDR), there were 95 DEGs tained mainly hypermethylated DMPs (NES > 0). Inter- (adjusted p-value < 0.1), of which 88 were upregulated estingly, the Slits2 protein has been described as an and 7 downregulated. The majority of differentially inhibitor of HIV-1 replication in T cells.35 When assess- expressed genes (DEGs, p-value < 0.05, limma model) ing correlations between DMPs and the total T-cell were protein coding genes (67%) and 7%, immunoglob- responses against entire HIV or HIVconsv measured by ulin coding genes (Table S4). In parallel, 5447 differen- IFNg ELISpot,13 the methylation levels of 32 CpG posi- tially methylated positions (DMPs, p-value < 0.05, tions showed a positive correlation with these parame- limma model, Table S5) were identified, that were sym- ters, while 22 were negatively correlated. Of relevance, metrically distributed in hypermethylated (n = 2864, the methylation levels of APOBEC3G, involved in HIV-1 Figure 1b red) and hypomethylated DMPs (n = 2583, antiviral function, and TAP2, a gene involved in antigen Figure 1b blue) one week after MVA.HIVconsv vaccina- presentation were both positively correlated with HIV tion, as shown in the volcano plot (Figure 1b and S2b). total and HIVconsv T-cell breadth and magnitude. Of After FDR adjustment only 2 hypermethylated DMPs the DMPs positively correlated with T-cell response remained significant (adjusted p-value < 0.1). Of the magnitude against HIVconsv but not total HIV, the identified DMPs (p-value < 0.05, limma model), 57% majority of associated genes were involved in chromatin mapped to regulatory regions (TSS200, TSS1500, regulation and gene transcription (e.g. MGMT, RNF20, 1stExon and 5’UTR regions), 38% to Body regions, and RABGGTB, TOX2 and CPEB.33,34 Finally, the methyla- 5% to 3’UTR regions (Figure S1c, and Table S6). Over- tion levels of AUTS2 gene, associated with the chroma- all, 40% of DMPs were located in CpG islands, while tin remodelling complex PRC1, were negatively 22% were from islands shores and 7%, from island correlated with the magnitude of the HIVconsv specific shelves. The remaining 32% of DMPs, were found to T-cell response.33,34 open sea regions (Figure S1d, and Table S6). These data demonstrate a strong impact of vaccina- While DEGs and DMPs between Vacc and baseline tion on the gene transcriptional program, particularly time points were widely distributed among the different for immunoglobulins genes and for immune response chromosomes (Figure 1c), the highest proportion of pathways including Fc receptors, interleukins and cyto- DEGs and DMPs were found in chromosome 1 (chr1), kines and complement signalling. The same interven- which is the largest chromosome. Overall, DEGs and tion impacted DNA methylation levels of genes DMPs in each chromosome were proportional to its involved in responses to infection. Interestingly, DNA length (Figure S1e, f, Tables S4 S6). In DNA methyla- methylation imprints correlated with measures of the tion, chr8 and chrX showed a significant reduction in HIVconsv vaccine-specific T cell immunity, suggesting the proportion of DMPs in comparison to CpG posi- a direct role of epigenetic modulation on vaccine tions that are included in the array (Figure S1e and responsiveness or an impact of vaccination on epige- Table S6). netic cascades. At pathway and blood transcription modules (BTMs) level, based on DEGs, the majority of genes that were upregulated after vaccination (Normalized Enrichment Cumulative transcriptional and methylation changes Score, NES > 0) were related to Fc receptors, comple- after vaccination and RMD intervention ment activation, phagocytosis and B cell receptor (BCR) RMD administration impacts the global chromatin con- activation. In parallel, the majority of downregulated densation state of host genome, thus affecting not only genes (NES < 0) were linked to pathways of GPCR (G HIV-1 transcription but also other host genes. To www.thelancet.com Vol 78 Month April, 2022 7 Articles Figure 1. Impact of vaccination on gene expression and DNA methylation in the BCN02 trial. (a) and (b) show the Volcano Plots for the 1287 DEGs and the 5447 DMPs between post-vaccination and baseline, respectively (p-value < 0.05). X-axis show the log2 Fold-Change, and Y-axis the -log10 p-value. Grey colour indicates p-value > 0.05; red, p-value < 0.05 and FC > 0 and blue, p- value < 0.05 and FC < 0. (c) CIRCOS plot that includes an outer track with the chromosomes, genes and CpG positions in the 8 www.thelancet.com Vol 78 Month April, 2022 Articles evaluate the cumulative effects of the intervention (Vacc (Transcription start site) and genic enhancers (EnhG2), +RMD) on the host transcriptome and epigenome, while hypomethylated DMPs dominated predominantly DEGs and DMPs between Vacc+RMD and baseline in regions of weak transcription (Figure 2d). Similarly, were identified (Tables S7, 8). Overall, 1404 genes were downregulated genes were found in regions of active upregulated and 1561 downregulated (Figure 2a, p-value TSS, and upregulated ones, in less active regions or in < 0.05, limma model), and after adjusting by false dis- regions associated with poised expression (Figure 2f). covery rate 240 genes were upregulated and 103, down- In addition, histone acetylation marks enrichment fol- regulated considering an adjusted p-value < 0.05. Of all lowed the same trend, with hypermethylated DMPs and DEG (p-value < 0.05, limma model), 79% were found downregulated DEGs being enriched in regions of in protein coding genes (Table S9). For DNA methyla- active transcription (e.g.H3K27ac) (Figure 2e and g). tion, there were 7025 hypermethylated and 9114 hypo- These results show that, although RMD is infused to methylated CpG positions between Vacc+RMD and increase histone acetylation, and consequently gene baseline considering a p-value < 0.05 (Figure 2b). After expression, one week after RMD infusion, the majority adjusting for multiple comparisons, 1301 DMPs were of upregulated genes were found in regions of transcrip- hypermethylated and 2707 hypomethylated (FDR tional repression or poised promoters and enhancers adjusted p-value < 0.05). Of the identified DMPs (p- rather than regions of active transcription. value < 0.05, limma model), 54% were located in regu- Since the Vacc+RMD intervention reflected the latory regions (5’UTR, 1stExon, TSS1500 and TSS200), cumulative changes induced by MVA.HIVconsv vacci- 42%, in body and, 4%, in 3’UTR (Figure S3a, Table nation and RMD infusion (Figure S2g, h), we identified S10). The number of DMPs in the regulatory regions RMD-specific effects, by conducting a transcriptomics 1stExon, TSS1500 and TSS200 were underrepresented assessment on samples from participants in the part A compared to their coverage in the array, while the pro- REDUC study (NCT02092116),6 in which ART-sup- portion of DMPs in 5’UTR and the body was slightly pressed participants received 3 cycles RMD without overrepresented (Table S10). Overall, 30.9% of the prior vaccination. As in BCN02, the gene expression DMPs were found in islands, 24.4%, in shores, 7.6%, profiles were compared between baseline and one week in shelves, and 37.1% in open sea regions, respectively after the 3-cycle RMD infusion (post-RMD). Due to lim- (Figure S3b, Table S10). The different DEGs and DMPs ited sample availability, no methylation analyses were were distributed in the different chromosomes, and conducted in these samples. However total PBMC and generally, the size of the chromosome was associated isolated CD4 T cell fractions were available for tran- with the number of DEGs and DMPs (Figure S3 c, d, scriptomic analysis, in which PBMCs showed 224 upre- Tables S9, 10). Only in chr3, a higher number of DMPs gulated and 567 downregulated genes, respectively (p = 0.01, Chi-square test) was found when compared to (Figure 2h, Table S11). When analysing purified CD4 their representation in the methylation array (Figure only, a higher number of DEGs was found, 883 upregu- S3d, Tables S9, 10). lated and 613 downregulated (Figure 2i, Table S12). In Of interest, RMD administration led to the specific both sample sets, DEGs were spread among all chromo- upregulation of DNMT3B and EZH2 enzymes (Figure somes (Figure 2j). Interestingly, the ChromHMM and S4), both key upstream regulators of DNA histone marks enrichment from CD4 DEGs in the methylation.36,37 Actually, RMD administration, as REDUC clinical trials (Figure 2k n) were similar to the other HDACi, impact epigenetic cascades at different results observed in BCN02 (Figure 2f, g). levels.38 To further define potential epigenetic landscape Restoration of dysregulated gene expression profiles changes from baseline to the Vacc+RMD time points, and their underlying epigenetic mechanisms may be enrichment analyses were applied on DMPs or DEGs critical to induce effective antiviral activity in kick-and- using the ChromHMM and Histone marks in PBMCs kill iinterventions. Therefore, gene set enrichment anal- from the Roadmap database (E062) (https://egg2.wustl. yses (GSEA) were conducted to identify the dysregu- edu/roadmap/web_portal/chr_state_learning.html) lated pathways upon Vacc+RMD (BCN02) or RMD only (Figure 2d g). These ChromHMM enrichment analy- treatment (REDUC) (Figure 3, Table S13). For the com- ses showed that hypermethylated DMPs at Vacc+RMD bined regimen of therapeutic vaccine plus RMD, the time point were enriched in regions of active TSS data show a strong impact on the HIV pathways in genome. The two inner tracks are the Manhattan plots for DMPs (green) and DEGs (orange) between post-vaccination and baseline time points. (d) and (f) show the pathway enrichment (GSEA) for DEGs and DMPs, respectively. Red colour reflects NES > 0 and blue, NES < 0. (e) Correlation plots of DEGs with breadth and magnitude of the virus vaccine-specific T-cell response evaluated by IFNg ELISpot as detailed in the methods section (p-value < 0.05 & Rho > |0.6|). For both, DEGs and magnitude and breadth of the T- cell response, the log2 change between vaccination and baseline was used. (g) Correlation plot for DMPs and T-cell responses (p- value < 0.01 & Rho > |0.6|) against the HIVconsv immunogen or total HIV-1. For both, DMPs and magnitude and breadth of the T cell response, the log2 change between vaccination and baseline was used. Blue indicates positive and red indicates negative Rho values (Spearman’s correlation). Blanks make reference to non-significant correlations. www.thelancet.com Vol 78 Month April, 2022 9 Articles Figure 2. Effect of combined intervention on gene expression and DNA methylation. (a) Volcano plot of the 3106 DEGs between time points Vacc+RMD and baseline. (b) Volcano plot of the 15139 DMPs between time points Vacc+RMD and baseline. (c) Manhattan plot in CIRCOS. The outer circle shows the chromosomes and positions of DEGs and DMPs. The inner tracks show the Manhattan plot for DMPs (green) and DEGs (orange) between Vacc+RMD and baseline time points. (d) shows the 18-state 10 www.thelancet.com Vol 78 Month April, 2022 Articles Figure 3. Effect of RMD on gene expression and DNA methylation in the BCN02 and REDUC trial. Enrichment Map of enriched pathways and BTMs (GSEA adjusted p-value < 0.2) in DEGs before and after RMD infusion in PBMCs of BCN02 (Orange) and in REDUC PBMC (Purple) or isolated REDUC CD4 T cells (Brown). DMPs before and after RMD infusion in BCN02 are shown (Green). The colour of the nodes indicates the different datasets. Edges represent the similarity of the nodes (cutoff = 0.7). infectious disease and HIV module (HIV infection and related processes, as well as the pathways DNA double HIV life cycle) as well as on the pathways in immune sig- strand break and cell cycle mitotic in modules DNA repair nature module (e.g. T cell activation and antigen presenta- and cell cycle. In parallel, the pathways in modules tion). Particularly, Host interactions of HIV factors, HIV transport and cell signalling were mainly impacted at infection, HIV life cycle, Infectious Disease and Cellular transcriptome level in isolated CD4 cells after RMD- responses to external stimuli, were enriched in DEGs of only treatment (REDUC). Finally, only few pathways REDUC CD4 cells after RMD only and in BCN02 showed an overlap between the two datasets in the mod- PBMC from individuals undergoing Vacc+RMD ule of immune signature, possibly due to the fact that (Figure 3). Interestingly, numerous other pathways the two studies differ by the inclusion of a therapeutic were shared between the two datasets, including path- vaccine in BCN02. Still, among the immune signature ways in the metabolism module like Metabolism of RNA module, pathways including signalling by interleukins, or TCA cycle and respiratory electron transport,) the path- FCERI mediated NF-kB activation and antigen processing way Translation from the protein module, the pathway cross presentation were dysregulated in isolated CD4 after Signalling by ROBO receptors in the module neuronal- RMD-only treatment. Interestingly, there was an ChromHMM enrichment and (e) the histone mark enrichment for DMPs. The same was done for DEGs in (f) and (g). (h) and (i) Vol- cano plots of for DEGs in total PBMCs and isolated CD4 T cell after RMD-only treatment (REDUC), respectively. Colour key is the same as in (a) and (b). (j) Manhattan plot for RMD-only treatment (REDUC) DEGs in PBMC (purple) and CD4 (brown). (k and m) ChromHMM 18-state enrichment based on DEGs in total PBMCs (in k) or CD4 T cells (in m). Histone marks enrichment for DEGs in PBMC (l) and in CD4 T cells (n). For Volcano plots (a, b, h, i) X-axis show the log2 Fold-Change, and Y-axis the -log10 p-value. Grey col- our indicates p-value > 0.05; red, p-value > 0.05 and FC > 0 and blue, p-value > 0.05 and FC < 0. For ChromHMM and Histone marks enrichments, the colour of the dots is associated with the log2 Fold-change enrichment, and the dot size, with the log10 p- value. www.thelancet.com Vol 78 Month April, 2022 11 Articles enrichment in DEGs from PBMC for pathways involv- with an Early versus Late rebound (DNA methylation ing B cell immunity, including B cell surface and AUC=0.812 and gene expression AUC=0.594 at the Enriched in B cells and Antigen activates BCR leading to Vacc+RMD time point, Figures 4a and S6a). The heat- generation of second messengers which were observed even maps in Figures 4b and S6b confirm these results. At in isolated CD4 T cells from the REDUC trial. Of note, earlier time points, the discriminating capacity of DNA the enrichment of the BCR signalling pathway in isolated methylation was lower (AUC BSL = 0.563, AUC CD4 T includes DEGs also present in the TCR activation Vacc = 0.687), which was also true for gene expression signalling. A similar observation was made for the Neuro- (AUC BSL = 0.563, AUC Vacc = 0.594). Importantly, nal-related process classification, where identified genes any clinical parameter, apart from time to viral rebound are not exclusively associated with processes in the cen- during MAP, was significantly different between Early tral nervous system but also includes molecules and Late rebounders (Table S14). involved in other process, including immunological syn- Given the potential of DNA methylation to discrimi- apses (GTPases, Calcium Voltage Gated channels, nate between Early and Late rebounders at the Vacc among others). In regards to the T cell related pathways, +RMD time point, the DMPs (p-value < 0.05, limma the majority of them were enriched in DEGs from the model) between Early and Late rebounders were further BCN02, although TCR signalling was also present in iso- analyzed. Of the identified DMPs, 3835 DMPs were lated CD4 from REDUC (Figure 3). Of note, aside from hypermethylated and 3761, hypomethylated in Early a large number of pathways in the gene expression rebounders; of which 5 and 1 DMPs, respectively, were module that were modulated by the combination ther- maintained significantly differentially methylated with apy, there was a marked regulation of pathways like a FDR adjusted p-value < 0.05 (Figure 4b, Table S15). HATs acetylated histones and Chromatin modifying The hypermethylated DMPs (p-value < 0.05, limma enzymes, indicating the complexity of the epigenetically model) in Early rebounders were found in bivalent and dysregulated mechanisms upon Vacc+RMD treatment inactive regions of the genome, which were enriched in and possibly reflecting RMD activity. the histone mark H3K27me3 (Figure 4c). In contrast, Focusing on the BCN02 dataset only (Figure S5), the hypomethylated DMPs in Early rebounders were there were multiple pathways suggesting an epigenetic found in active regions of the genome. Interestingly, regulation, as they showed opposite NES. Many of these when assessing DMPs at transcription factor binding pathways located to the immune signature module, like sites (TFBS), hypermethylated DMPs in Early rebound- Enriched in Monocytes, which show a majority of upregu- ers were strongly enriched for factors of the ETS family, lated and hypomethylated genes. Also, T cells and T cell while the hypomethylated DMPs were enriched in tran- activation, mainly contained hypermethylated DMPs scription factors of the Zf, bZIP and IRF families and downregulated expression of genes. The same trend (Figure 4d, e). These data suggest that the epigenetic was observed for HIV life cycle pathway and in other landscape after vaccination and RMD infusion may aid pathways of the modules cell cycle, metabolism, gene in controlling viral replication either directly, by the expression or protein translation. Overall these results modulation of chromatin accessibility in regions of suggest that after a Vacc+RMD regimen, a homeostatic active/inactive transcription, or through the binding of mechanism might be activated as the global epigenetic specific TFs. state of the cells attempts to counteract the acetylation When analysing the main pathways enriched by effects induced by RMD, whose main effects (4 h post DMPs between Early and Late rebounders at the Vacc infusion) waned 1 week after its administration. +RMD time point, differential methylation levels were observed in genes involved in pathways and modules related to HIV host factors, immune parameters (Anti- DNA methylation imprints after intervention can gen presentation, TCR signalling and signalling by interleu- predict viral rebound before ATI kins), metabolism of lipids and RNA, membrane Next, the hypothesis that changes induced by vaccina- trafficking and ER-Golgi transport, MAPK and TGF-B tion and RMD treatment before MAP could explain the signalling, and cell cycle processes (Figure S7). To fur- different viral rebound kinetics was tested. An unsuper- ther characterize these pathways, heatmaps of pathway- vised PCA of DNA methylation data in samples from associated DMPs (p-value < 0.01 and log2FC > |0.5|, the Vacc+RMD time point showed individuals with limma model) were created and the correlation with early or late rebound (VL reaching 2000 HIV RNA cop- viral and immune parameters was studied (Figure 4b ies/mL before [“Early rebounders”] or after [“Late and Table S16). These analyses showed that hypomethy- rebounders”] 4 weeks of MAP) to cluster closely lated CpGs in Late rebounders and thereby, hyperme- together (Figure 4a). The same clustering was not thylated in Early rebounders, positively correlated with observed when performing PCA with gene expression three measures of viral reservoir and replicative activity data (Figure S6a). A logistic regression model using the under ART, including HIV proviral levels, plasma viral PC1 and PC2 as covariates, confirmed the potential of load and cell-associate RNA levels (CA-RNA) as well as DNA methylation to discriminate between individuals with the magnitude of the IFNg T cell responses against 12 www.thelancet.com Vol 78 Month April, 2022 Articles Figure 4. Differential DNA methylation between individuals with an Early or Late rebound. (a) PCA for DNA methylation at time point Vacc+RMD. Blue and orange colours indicate Early and Late rebounders, respectively. (b) Heatmap of Z-score methylation levels of DMPs (p-value < 0.01) between Early and Late rebound at Vacc+RMD in different pathways (Y-axis). The right part of the heatmap shows the correlation (Spearman’s Rho) between DMPs methylation levels and different parameters. For correlations with www.thelancet.com Vol 78 Month April, 2022 13 Articles HIVconsv and HIVtotal (HIV regions not covered by the correlated with the time of undetectable viral load in immunogen) one week before the second vaccination. MAP, while methylation levels in JAK1 were signifi- On the other hand, hypermethylated DMPs in Early cantly related to the magnitude of the T cell response individuals were negatively correlated with the time off against HIV (Figure S8e, f, h). Next, for the metabolism antiretroviral treatment (time MAP) and the time with module there were 9 DMPs associated with the RUFY1 undetectable viral load during MAP (UD MAP). In addi- gene, which was positively correlated to the time on tion, these DMPs showed an association with the time cART at BCN02 entry (Figure S8g, h). Interestingly, the from HIV infection until cART initiation (HIV to cART) levels of gene associated DMPs classed in metabolism, as well as the time on cART at BCN02 entry (Time on cell cycle and cell transport with reported interaction cART at BCN02). Although DNA methylation levels with HIV virus (HADHA,42,43 KIF2A,44 CHMP2B44) were not assessed at time points in early and untreated correlated with viral and immune parameters HIV infection, these data suggest that some epigenetic (Table S16). marks could be defined even before ART initiation and Of special importance, of the 150 DMPs that differed that the intrinsic individual epigenome or the impact of between Early and Late rebounders at the Vacc+RMD the virus during acute infection might define the ability time point and represented in the GSEA analysis to control HIV-1. This is in line with a recent study that (Figure S7), 66 were also differentially methylated demonstrated certain epigenetic changes induced upon between the two groups at the baseline time point and acute HIV-1 infection not to revert upon ART.39 maintained a differential methylation signature after Focusing on specific DMPs in the immune signature vaccination. RMD administration further increased the module (Figures 4b and S8), the methylation levels of difference of the median methylation levels between the PRKCQ gene were negatively correlated with cell-associ- two groups for 30 of these 66 DMPs. In addition, 52 of ated (CA)-RNA and proviral levels (Figure S8a, h). Inter- the 66 DMPs were also differentially methylated estingly, this gene has been shown to be relevant for between the two groups in MAP time point. Such HIV-1 latency reversal.40,41 Differential methylation was results indicate that multiple DMPs between Early and also observed for genes in the HLA locus. In particular, Late rebounders are maintained during MAP and in the DNA methylation in positions of the HLA-DPA1 and presence of viral replication (Table S15). HLA-DRB1 loci were negatively correlated with days off Together, these results provide evidence that specific cART parameter (Figure S8b, c, h). In the case of host-gene methylation profiles are associated with clini- MAP2K3, a gene annotated in the modules HIV and cal variables (time on MAP, time on ART, time on immune signature, and hypermethylated in Late HIV), virological parameters (viral loads and HIV provi- rebounders, its methylation levels are negatively corre- ral levels) and immunological determinations (T-cell lated with viral load while a positive correlation is responses against HIV and HIVconsv immunogen observed with the time from HIV infection to cART measured by IFNg ELISpot) and allow for the discrimi- treatment initiation, the time with undetectable viremia nation between Early and Late rebounders in a kick- in MAP and the time off cART (Figure S8d, h). On the and-kill therapeutic vaccination strategy. Additionally, other hand, in these two categories, the IL6 and JAK1 the data also show that RMD impacts the epigenetic genes were hypomethylated in Late rebounders, with landscape broadly and gradually accentuates the early methylation levels positively correlated with CA-RNA. differences in DNA methylation between individuals In addition, methylation levels in IL6 positively with an Early or Late virus rebound. histone 3 acetylation (H3Ac), proviral levels, ultrasensitive viral load and cell-associated RNA (CA-RNA), red colour shows a negative correlation and blue a positive one. For correlations with the magnitude and breadth of the HIVconsv- or HIVtotal-specifc T cell response, green shows a negative correlation and yellow a positive one. For correlations with viral load at day of cART resumption (VL MAP), time on MAP and time in MAP with undetectable viral load (UD MAP), orange shows a negative correlation and brown a positive one. For correlations with VL at cART initiation, time from HIV infection until cART initiation (HIV to cART) and time on cART at BCN02 entry, grey shows a negative correlation and purple a positive one. (c) 18-state ChromHMM enrichment and (d) the his- tone marks enrichment for DMPs between and Early and Late rebounders. In the x axis, Hyper and Hypo refer to hypermethylated or hypomethylated DMPs in Early rebounding individuals. The Log2 Fold-Change is shown by the dot colour, while the -log10 p-value, with the dot size. (e) Transcription factor (TF) enrichment based on DMPs hypermethylated in Early individuals, and (f) based on DMPs hypomethylated in Early individuals. Colours indicate the TF family, and the name of the TF is shown when Log2 Fold-Change > 0.3. (F) Heatmap of Z-score methylation levels of DMPs (p-value < 0.01) between Early and Late rebound at Vacc+RMD in different pathways (Y-axis). The right part of the heatmap shows the correlation (Spearman’s Rho) between DMPs methylation levels and dif- ferent parameters. For correlations with histone 3 acetylation (H3Ac), proviral levels, ultrasensitive viral load and cell-associated RNA (CA-RNA), red colour shows a negative correlation and blue a positive one. For correlations with the magnitude and breadth of the HIVconsv- or HIVtotal-specifc T cell response, green shows a negative correlation and yellow a positive one. For correlations with viral load at day of cART resumption (VL MAP), time on MAP and time in MAP with undetectable viral load (UD MAP), orange shows a negative correlation and brown a positive one. For correlations with VL at cART initiation, time from HIV infection until cART initia- tion (HIV to cART) and time on cART at BCN02 entry, grey shows a negative correlation and purple a positive one. 14 www.thelancet.com Vol 78 Month April, 2022 Articles Discussion recall immune response, as the individuals in BCN02 This study assessed the impact of a kick-and-kill strategy were previously vaccinated in BCN01 with an heterolo- combining therapeutic T-cell vaccination and RMD on gous vector prime-boost of ChAd63.HIVconsv and host transcriptional and epigenomic patterns. It also MVA.HIVconsv.15 Thus, the selected time points could tested the hypothesis that epigenetic imprints could blur the observation of a major effect on immune tran- serve as biomarkers of HIV control during the moni- scriptional effects. At the same time, epigenetic signa- tored antiretroviral pause (MAP). BCN02 was a phase I, tures, which are more stable over time than open-label clinical trial that combined the latency revers- transcriptomic changes54,55 may be less sensitive to pre- ing agent (LRA) romidepsin (RMD), an inhibitor of his- cise timing. tone deacetylases (HDACi), with HIVconsv vaccination The availability of samples from the REDUC trial, in order to reactivate the HIV provirus at the time of a where participants received RMD without a therapeutic boosted T-cell response against conserved parts of the T-cell vaccine, allowed to assess the contribution of virus. During the MAP, 4 out of 13 individuals main- RMD to the combined intervention in BCN02. Indeed, tained a VL < 2000 HIV RNA copies/mL for more than this data show that RMD administration alone had a 4 weeks.13 Systems biology analyses identified molecu- marked effect on both, gene expression and DNA meth- lar pathways, including those involved in infectious dis- ylation, although at gene-level, no major negative corre- eases and immune activation, that were modulated lations were captured among them. This outcome is during the combined intervention at transcriptional and likely partially due to the low sample size and the differ- epigenetic levels. Finally, differentially DNA methylated ent time to respond to stimuli. Nevertheless, different positions (DMPs) after the intervention (Vacc+RMD) DMPs and DEGs converged in the same pathways, sug- discriminated individuals with an Early or Late virus gesting an epigenetic regulation of these pathways. rebound during MAP, with some signatures having Among the pathways modulated by Vacc+RMD, path- been present even before the therapeutic intervention. ways related to HIV life cycle and infectious diseases The present analysis shows that MVA.HIVconsv vac- were identified as the most severely impacted. The fact cination had a marked effect on the expression levels of that these observations were made using samples taken genes involved in the humoral immune response (e.g. after the most prominent effect of RMD on histone acet- IGHG1, IGHV1-24 and C1QA), as well as on genes ylation had already waned, suggests that a dysregulation involved in interleukin signalling. After vaccination, of HIV transcription is further maintained. In fact, the there was also an impact on DNA methylation levels of REDUC trial showed that while RMD was able to induce genes involved in infectious disease pathways (e.g. the initiation and elongation of the virus transcription, APOBEC3G, TUBB3, and FURIN) and, importantly, on the drug failed to increase late events in HIV-1 transcrip- genes potentially involved in further downstream epige- tion.56 This finding is in line with several other reports netic processes, including PRC1, MGMT and RNF20. that suggest that more potent LRA, and potentially com- Interestingly, the methylation levels of genes associated binations of them would be needed to efficiently purge with epigenetic processes were directly correlated with the viral reservoir. Combining LRA could indeed allow the magnitude of the vaccine-elicited T-cell response for broader reactivation of latent virus through activity specific for the conserved HIV regions.45 49 This obser- on different distinct integration sites in the host vation is in line with a previous report on influenza vac- genome and in different cell types.57,58 cination, demonstrating that gene-associated DMPs In addition to a strong latency reversing activity that correlated with the vaccine-induced humoral is able to induce HIV antigen expression, the stimula- response.48,50 Similarly, BCG vaccination has been tion of an effective immune response to eliminate cells reported to induce an innate immune like-memory in with reactivated virus will also be critical to achieve monocytes that was driven by epigenetic processes, and HIV-1 remission in the absence of ART.8,59 In the some reports suggest this antigen-unspecific effect may BCN02 study, the virus-specific T-cell immunity was also be induced by other vaccine vectors and clearly enhanced during the therapy13,60 and was found adjuvants.48,51,52 However, in all these analyses, choos- here to relate to pathways (including T-cell activation, ing the adequate time point to document possible asso- TLR signalling and/or antigen presentation) that were ciations between immune response to vaccination and modulated at both, the epigenetic and transcriptional transcriptomic changes or epigenetic profiles appeared levels. These pathways were also observed in the tran- to be critical. This dependence on time point also scriptional program of CD4 cells from the individuals applies to the present study, were the selection of other in the REDUC clinical trial, in which only RMD was time points than the one studied (1 week after the first administered. These findings suggest that the vaccina- BCN02 MVA.HIVconsv vaccination) might have helped tion of individuals in BCN01 and BCN02 prior to RMD to capture more profound effects on the vaccine- infusion may have poised the immune system to induced T cell response and allowed to better document respond and exert its antiviral activity against HIV. the transient changes in transcriptomic activity.53 In Indeed, in vitro studies showed that inducing an effi- addition, the observed transcriptomics patterns reflect a cient HIV-specific CD8 response before the www.thelancet.com Vol 78 Month April, 2022 15 Articles administration of a LRA resulted in a more effective proviral and CA-RNA levels (Figure S9), thus further removal of infected cells.8 suggesting that the hosts epigenetic landscape may be a Genes in pathways of HIV infection, immune signa- critical determinant for viral integration. ture and metabolism categories, among others, showed The limitations of this work include small study size, a differential DNA methylation between individuals limited sampling time points and lack of a placebo con- with an Early and Late viral rebound. Interestingly, trol arm. This prevented us from determining, for some of these methylation imprints were already example, to what extent the basal levels of DNA methyl- observed at baseline could thus be host-specific baseline ation (BCN01 enrolment) could have predicted viral characteristics or, alternatively, be modulated by HIV rebound kinetics during the treatment interruption infection, antiretroviral treatment or previous prime- phase. Actually, in well powered studied selection of a boost vaccination in BCN01. Thus, the present data hint combination of DNA methylation candidates and viral at the possibility that epigenetic signatures, including and clinical data may yield with predictive models of those assessed before kick-and-kill interventions, could viral rebound. In addition, the coverage of the methyla- predict eventual virus control in treatment interruption. tion studies could be increased by using the EPIC Bead- Furthermore, a recent study has associated specific chip array rather than the 450K Beadchip array, as the DNA methylation marks with the size of the viral reser- former includes additional DMPs on regulatory ele- voir and the time to viral rebound during an ATI in a ments that may be relevant to understand the differen- clinical trial administrating a latent reactivation agent tial viral rebound. Another limitation is the unbalanced (NCT01680094).61 The potential of epigenome studies sex distribution in this study with only one female par- in the identification of correlates of control of HIV-1 ticipant. This prevented us from applying statistical cor- infection is also in line with a previous report from our rection methods and, given the distinct clustering of group showing that differential capacity to control HIV this individual, obliged us to exclude this participant infection in the absence of ART was associated with a from the omics analyses. Furthermore, and although differential methylation profile of genes involved in HIVconsv was designed to elicit T-cell responses, the antiviral response and T-cell immunity.16,62 Finally, dif- gene expression signature one week after vaccination ferent DNA methylation in CD4 T cells has been associ- included markers of immunoglobulins, Fc receptors ated with distinct viral progression and can be impacted and complement activation. This suggests that vaccina- by cART treatment initiation,17 again suggesting tion may have driven some humoral immune activation, that epigenetic profiling, even in early stages of HIV including some that could target the viral vector MVA. infection, may have predictive power for post-treatment Unfortunately, no B cell immune-monitoring was control. included in BCN02, although this should be considered Previous report in Cutaneous T Cell Lymphoma in future clinical interventions. Another potential limi- showed that RMD increased the chromatin accessibility tation was the study of total PBMCs rather than sorted mainly of regions with open or relaxed chromatin,63 lymphocyte subsets, as it is not possible to determine while the present data showed that RMD administration the precise cellular origin of some of the observed sig- further potentiated the epigenetic differences between nals. To overcome this limitation, samples from the individuals with an early or late viral rebound. Such REDUC trial were evaluated and showed different sig- observations were not present at transcriptional level nals from PBMC or from isolated CD4 T cells. Interest- (Figure S4B, C). Interestingly, at Vacc+RMD time point, ingly, the transcriptional patterns in the isolated CD4+ Early rebounders showed hypomethylation mainly in T cells were similar to those observed in PBMCs from regions of active transcription, while Late rebounders the BCN02 study. With this result, it is tempting to were hypomethylated in regions of repressed chroma- speculate that previous vaccination may increase the sig- tin. This differential epigenetic landscape could also nals from CD4 T cells in BCN02 clinical trial, although affect viral rebound, since late rebounding individuals this should be tested in CD4 T cells isolated fractions may have HIV-1 integrated in regions of the host from BCN02 study, which are unfortunately unavail- genome that are transcriptionally less active, as has able. Therefore, although from a biomarker point of been suggested in the setting of natural HIV control.64 view, the use of total PBMCs is potentially still appeal- In this line, effects on DNA methylation were ing, for the identification of possible therapeutic targets, observed in CpG positions of a common integration isolated cell types studies might be more informative. site, BACH265 (Figure S9). Specifically, the Together, this is a unique exploratory study of a clini- cg03035849_BACH2 site was hypermethylated in Early cal trial testing a kick-and-kill strategy that showed a rebounders compared with Late rebounders one week partial clinical signal in terms of virus control during after RMD administration and positively correlated with treatment interruption. The data clarified the different CA-RNA. Similarly, two recently described integration molecular pathways that were modulated in response to sites, RASA3 and RPTOR,66 68 were differentially this specific intervention and highlighted that DNA methylated between Early and Late rebounders after methylation profiling could be further explored as a bio- romidepsin administration, and also correlated with marker to predict viral control before treatment 16 www.thelancet.com Vol 78 Month April, 2022 Articles interruption. These results warrant further confirma- funding from the European Union Horizon 2020 tion in larger, placebo-controlled clinical trials and may Framework Programme for Research and Innovation inform future refined strategies to achieve a functional under Grant Agreement N° 681137-EAVI2020 and N° HIV cure. 847943-MISTRAL, the Ministerio de Ciencia e Innovacion (SAF2017_89726_R), and the National Institutes of Health National Institute of Allergy and Contributors Infectious Diseases Program Grant P01-AI131568. This MRR and CB designed the experimental plan. BM, JM work was also supported in part by Grifols, a research and CB contributed with clinical trial, patient data and agreement with Aelix Therapeutics S,L and the Founda- sample management. BCl, BM, JM, OS, MM, were in tion Dormeur (Vadur, Liechtenstein. The BCN02 study charge of patient recruitment and sample provision. was funded in part by the ISCIII PI15/01188 grant, the BOT, CDC, SC and ALL did the sample processing. HIVACAT Catalan research program for an HIV vac- BOT, AEC and EG performed the transcriptome analy- cine and the Fundacio Gloria Soler. J.M-P. research pro- sis and BOT, MRR and FCM analyzed methylome data gram is funded by grant PID2019-109870RB-I00 from and their integration with clinical and virological the Spanish Ministry of Science and Innovation. MLC parameters and immune reactivity data. ME and MB was partially supported by the Spanish Ministry of performed DNA Methylation assays and helped in data Economy, Industry and Competitiveness, Reference analysis. BM, ALL, MRU and SC assessed HIV specific PID2019-104830RB-I00. AE-C was funded by ISCIII T cell responses. MCP and JMP determined CA-HIV-1 /MINECO (PT17/0009/0019) and co-funded by DNA, CA-HIV-1 RNA and residual viral load. BOT con- FEDER. This work was carried out within the frame- ducted R language and programing for the execution of work of the PhD in Advanced Immunology of the Uni- statistically, bioinformatic and integrative analyses with versitat Autonoma de Barcelona for BO-T and CD-C. support of AEC, AS and MLC. BOT, MRR, and CB did BO-T received an EMBO short term fellowship. the results interpretation. BOT, MRR, CB, BC, BM, MT, TH, DH, and RP participate in scientific discussions. BOT, CB and MRR drafted and edited the manuscript Supplementary materials together. All authors reviewed and approved the final Supplementary material associated with this article can version of the manuscript. be found in the online version at doi:10.1016/j. ebiom.2022.103956. Data sharing statement All gene expression and DNA methylation data were References 1 Ward AR, Mota TM, Jones RB. Immunological approaches to HIV uploaded in GEO under accession numbers: cure. Semin Immunol. 2021;51: 101412. Available from; https://link- GSE184653, GSE185391 and GSE185027, which include inghub.elsevier.com/retrieve/pii/S1044532320300282. sample basic information including sample source, age, 2 Ndung’u T, McCune JM, Deeks SG. Why and where an HIV cure is needed and how it might be achieved. Nature. 2019;576(7787):397– sex, time point and viral rebound. Additional informa- 405. Available from; https://www.nature.com/articles/s41586-019- tion can be found in the supplementary material of this 1841-8. 3 Bailon L, Mothe B, Berman L, Brander C. Novel approaches article. Additionally, the R code applied in the analyses towards a functional cure of HIV/AIDS. Drugs. 2020;80:859–868 of this article can be found in the GitHub link: https:// Available from; https://doi.org/10.1007/s40265-020-01322-y. github.com/hostimmuneOMICS/BCN02_OmicsAnaly 4 Kim Y, Anderson JL, Lewin SR. Getting the “kill” into “shock and kill”: strategies to eliminate latent HIV. Cell Host Microbe. sis. 2018;23:14–26. Cell Press. Available from; https://doi.org/ 10.1016/j.chom.2017.12.004. 5 Delagreverie HM, Delaugerre C, Lewin SR, Deeks SG, Li JZ. Ongo- ing clinical trials of human immunodeficiency virus latency-revers- Declaration of interests ing and immunomodulatory agents. Open Forum Infect Dis. 2016;3 BM is a consultant of AELIX THERAPEUTICS, S.L out- (4). Available from; https://academic.oup.com/ofid/article/3/4/ ofw189/2593325. side the submitted work. CB is co-founder, chief science 6 Søgaard OS, Graversen ME, Leth S, et al. The depsipeptide romi- officer and shareholder of AELIX THERAPEUTICS. J. depsin reverses HIV-1 latency in vivo. Siliciano R.F., editor PLOS M. has received research funding, consultancy fees and Pathog. 2015;11:(9) e1005142. Available from; https://doi.org/ 10.1371/journal.ppat.1005142. lecture sponsorships from and has served on advisory 7 Elliott JH, Wightman F, Solomon A, et al. Activation of HIV tran- boards for various companies (MSD, Gilead Sciences, scription with short-course vorinostat in HIV-infected patients on suppressive antiretroviral therapy. Cullen B.R., editor PLoS Pathog. Viiv Healthcare, and Janssen-Cilag). All other authors 2014;10:(11) e1004473. Available from; https://doi.org/10.1371/ declare that they have no competing interests. journal.ppat.1004473. 8 Shan L, Deng K, Shroff NS, et al. Stimulation of HIV-1-specific cytolytic T lymphocytes facilitates elimination of latent viral reser- voir after virus reactivation. Immunity. 2012;36(3):491–501. Avail- Acknowledgments able from; http://linkinghub.elsevier.com/retrieve/pii/ S1074761312000842. The authors appreciate the participation of all the volun- 9 Ruiz A, Blanch-Lombarte O, Jimenez-Moyano E, et al. Antigen pro- teers in this study. This publication has received duction after latency reversal and expression of inhibitory receptors www.thelancet.com Vol 78 Month April, 2022 17 Articles in CD8+ T cells limit the killing of HIV-1 reactivated cells. Front 27 Merico D, Isserlin R, Stueker O, Emili A, Bader GD. Enrichment Immunol. 2019;9:3162. Available from; https://www.frontiersin. map: a network-based method for gene-set enrichment visualiza- org/article/10.3389/fimmu.2018.03162/full. tion and interpretation. PLoS One. 2010;5(11). Available from; 10 Sereti I, Krebs SJ, Phanuphak N, et al. Persistent, albeit reduced, https://pubmed.ncbi.nlm.nih.gov/21085593/. chronic inflammation in persons starting antiretroviral therapy in 28 Li S, Rouphael N, Duraisingham S, et al. Molecular signatures of acute HIV infection. Clin Infect Dis. 2017;64(2):124–131. Available antibody responses derived from a systems biology study of five from; https://academic.oup.com/cid/article-lookup/doi/10.1093/ human vaccines. Nat Immunol. 2014;15(2):195–204. Available cid/ciw683. from; https://pubmed.ncbi.nlm.nih.gov/24336226/. 11 Saez-Cirion A, Bacchus C, Hocqueloux L, et al. Post-treatment 29 Kucera M, Isserlin R, Arkhangorodsky A, Bader GD. AutoAnno- HIV-1 controllers with a long-term virological remission after the tate: a Cytoscape app for summarizing networks with semantic interruption of early initiated antiretroviral therapy ANRS VIS- annotations [version 1; referees: 2 approved]. F1000Research. CONTI study. PLoS Pathog. 2013;9(3). Available from; https:// 2016;5. Available from: /pmc/articles/PMC5082607/ . pubmed.ncbi.nlm.nih.gov/23516360/. 30 Sheffield NC, Bock CL. Enrichment analysis for genomic region 12 Fidler S, St€ohr W, Pace M, et al. Antiretroviral therapy alone versus sets and regulatory elements in R and Bioconductor. Bioinformatics. antiretroviral therapy with a kick and kill approach, on measures of 2016;32(4):587–589. Available from; https://academic.oup.com/ the HIV reservoir in participants with recent HIV infection (the bioinformatics/article/32/4/587/1743969. RIVER trial): a phase 2, randomised trial. Lancet. 2020;395 31 Heinz S, Benner C, Spann N, et al. Simple combinations of line- (10227):888–898. Available from; https://pubmed.ncbi.nlm.nih. age-determining transcription factors prime cis-regulatory ele- gov/32085823/. ments required for macrophage and B cell identities. Mol Cell. 13 Mothe B, Rosas-Umbert M, Coll P, et al. HIVconsv vaccines and 2010;38(4):576–589. Available from; https://pubmed.ncbi.nlm. romidepsin in early-treated HIV-1-infected individuals: safety, nih.gov/20513432/. immunogenicity and effect on the viral reservoir (Study BCN02). 32 Gu Z, Eils R, Schlesner M. Complex heatmaps reveal patterns and Front Immunol. 2020;11(10):823. Available from; https://pubmed. correlations in multidimensional genomic data. Bioinformatics. ncbi.nlm.nih.gov/32435247/. 2016;32(18):2847–2849. Available from; https://pubmed.ncbi. 14 Letourneau S, Im EJ, Mashishi T, et al. Design and pre-clinical nlm.nih.gov/27207943/. evaluation of a universal HIV-1 vaccine. PLoS One. 2007;2(10): 33 Stelzer G, Rosen N, Plaschkes I, et al. The GeneCards suite: from e984. Available from; https://journals.plos.org/plosone/article? gene data mining to disease genome sequence analyses. Current id=10.1371/journal.pone.0000984. Protocols in Bioinformatics20162016:1. Jun 1201630.1-1.30.33. Avail- 15 Mothe B, Manzardo C, Sanchez-Bernabeu A, et al. Therapeutic vac- able from; https://currentprotocols.onlinelibrary.wiley.com/doi/ cination refocuses T-cell responses towards conserved regions of full/10.1002/cpbi.5. HIV-1 in early treated individuals (BCN 01 study). EClinicalMedi- 34 GeneCards the human gene database [Internet]. www.genecards. cine. 2019;11:65–80. Available from; https://pubmed.ncbi.nlm.nih. org. 2017. Available from: https://www.genecards.org/. gov/31312806/. 35 Anand AR, Nagaraja T, Ganju RK. A novel role for Slit2/Robo1 axis 16 Oriol-Tordera B, Berdasco M, Llano A, et al. Methylation regulation in modulating HIV-1 replication in T cells. AIDS. 2011;25 of Antiviral host factors, interferon stimulated genes (ISGs) and T- (17):2105–2111. Available from: /pmc/articles/PMC4123625/ . cell responses associated with natural HIV control. Douek DC, edi- 36 Gagliardi M, Strazzullo M, Matarazzo MR. DNMT3B functions: tor PLOS Pathog. 2020;16:(8) e1008678. Available from; https:// novel insights from human disease. Front Cell Dev Biol. dx.plos.org/10.1371/journal.ppat.1008678. 2018;6:140. Oct 22OCT. 17 Moron-Lopez S, Urrea V, Dalmau J, et al. The genome-wide meth- 37 Vire E, Brenner C, Deplus R, et al. The Polycomb group protein ylation profile of CD4+ T cells from individuals with human EZH2 directly controls DNA methylation. Nat. 2005;439 immunodeficiency virus (HIV) identifies distinct patterns associ- (7078):871–874. 2005 4397078. Available from; https://www. ated with disease progression. Clin Infect Dis. 2020. Available nature.com/articles/nature04431. from; https://pubmed.ncbi.nlm.nih.gov/32712664/. 38 Rollins B, Martin MV, Morgan L, Vawter MP. Analysis of whole 18 Dobin A, Davis CA, Schlesinger F, et al. STAR: ultrafast universal genome biomarker expression in blood and brain. Am J Med Genet RNA-seq aligner. Bioinformatics. 2013;29(1):15–21. Available from; B Neuropsychiatr Genet. 2010;153B(4):919. Available from: /pmc/ https://academic.oup.com/bioinformatics/article/29/1/15/272537. articles/PMC3098564/ . 19 Li B, Dewey CN. RSEM: accurate transcript quantification from 39 Corley MJ, Sacdalan C, Pang APS, et al. Abrupt and altered cell- RNA-Seq data with or without a reference genome. BMC Bioinform. type specific DNA methylation profiles in blood during acute HIV 2011;12(1):1–16. Available from; http://www.biomedcentral.com/ infection persists despite prompt initiation of ART. PLOS Pathog. 1471-2105/12/323. 2021;17:(8) e1009785. Available from; https://journals.plos.org/ 20 Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor plospathogens/article?id=10.1371/journal.ppat.1009785. package for differential expression analysis of digital gene expres- 40 Lopez-Huertas MR, Mateos E, Dıaz-Gil G, et al. Protein kinase Cu sion data. Bioinformatics. 2009;26(1):139–140. Available from; is a specific target for inhibition of the HIV type 1 replication in https://pubmed.ncbi.nlm.nih.gov/19910308/. CD4 + T lymphocytes. J Biol Chem. 2011;286(31):27363–27377. 21 Ritchie ME, Phipson B, Wu D, et al. Limma powers differential Available from; http://www.jbc.org/article/S0021925820503426/ expression analyses for RNA-sequencing and microarray studies. fulltext. Nucleic Acids Res. 2015;43(7):e47. Available from; https://academic. 41 Lopez-Huertas MR, Li J, Zafar A, et al. PKCu and HIV-1 transcrip- oup.com/nar/article/43/7/e47/2414268. tional regulator Tat co-exist at the LTR promoter in CD4+ T cells. 22 Aryee MJ, Jaffe AE, Corrada-Bravo H, et al. Minfi: a flexible and Front Immunol. 2016;7:69. FEB. comprehensive bioconductor package for the analysis of Infinium 42 J€ager S, Cimermancic P, Gulbahce N, et al. Global landscape of DNA methylation microarrays. Bioinformatics. 2014;30(10):1363– HIV-human protein complexes. Nature. 2011;481(7381):365–370. 1369. Available from; http://bioconductor.org/packages/release/ Available from; https://pubmed.ncbi.nlm.nih.gov/22190034/. bioc/html/minfi.html. 43 Luo Y, Jacobs EY, Greco TM, et al. HIV-host interactome revealed 23 Fortin JP, Triche TJ, Hansen KD. Preprocessing, normalization directly from infected cells. Nat Microbiol. 2016;1(7):16068. Jul 23. and integration of the Illumina HumanMethylationEPIC array Available from; https://pubmed.ncbi.nlm.nih.gov/27375898/. with minfi. Bioinformatics. 2017;33(4):558–560. Available from; 44 Jarboui MA, Bidoia C, Woods E, et al. Nucleolar protein trafficking https://academic.oup.com/bioinformatics/article/33/4/558/ in response to HIV-1 Tat: rewiring the nucleolus. PLoS One. 2012;7 2666344. (11):e48702. Available from; https://pubmed.ncbi.nlm.nih.gov/ 24 Morris TJ, Butcher LM, Feber A, et al. ChAMP: 450k chip analysis 23166591/. methylation pipeline. Bioinformatics. 2014;30(3):428–430. Available 45 Cirovic B, de Bree LCJ, Groh L, et al. BCG vaccination in humans from; www.illumina.com. elicits trained immunity via the hematopoietic progenitor compart- 25 Leek T, Johnson J, Evan Parker W. SVA: surrogate variable analysis ment. Cell Host Microbe. 2020;28(2):322–334. Available from: [Internet]. 2019. Available from: https://bioconductor.org/pack /pmc/articles/PMC7295478/ . ages/release/bioc/html/sva.html. 46 Lai AY, Mav D, Shah R, et al. DNA methylation profiling in 26 Reimand J, Isserlin R, Voisin V, et al. Pathway enrichment analysis human B cells reveals immune regulatory elements and epige- and visualization of omics data using g:Profiler, GSEA, Cytoscape netic plasticity at Alu elements during b-cell activation. Genome and EnrichmentMap. Nat Protoc. 2019;14(2):482–517. Available Res. 2013;23(12):2030–2041. Available from: /pmc/articles/ from; https://www.nature.com/articles/s41596-018-0103-9. PMC3847773/ . 18 www.thelancet.com Vol 78 Month April, 2022 Articles 47 Dogra P, Ghoneim HE, Abdelsamed HA, Youngblood B. Generat- 58 Grau-Exposito J, Luque-Ballesteros L, Navarro J, et al. Latency ing long-lived CD8+ T-cell memory: insights from epigenetic pro- reversal agents affect differently the latent reservoir present in dis- grams. Eur J Immunol. 2016;46(7):1548–1562. Available from: tinct CD4+ t subpopulations. PLoS Pathog. 2019;15:(8) e1007991 /pmc/articles/PMC5035674/ . Available from; https://doi.org/10.1371/journal.ppat.1007991. 48 Bannister S, Messina NL, Novakovic B, Curtis N. The emerging 59 Ruiz A, Blanch-Lombarte O, Jimenez-Moyano E, et al. Antigen pro- role of epigenetics in the immune response to vaccination and duction after latency reversal and expression of inhibitory receptors infection: a systematic review. Epigenetics. 2020;15:555–593. Taylor in CD8+ T cells limit the killing of HIV-1 reactivated cells. Front and Francis Inc. Available from: /pmc/articles/PMC7574386/ . Immunol. 2019;10:3162. Available from; www.frontiersin.org. 49 Letourneau S, Im EJ, Mashishi T, et al. Design and pre-clinical 60 Rosas-Umbert M, Ruiz-Riol M, Fernandez MA, et al. In vivo effects evaluation of a universal HIV-1 vaccine. Nixon D, editor PLoS One. of romidepsin on T-cell activation, apoptosis and function in the 2007;2(10):e984. Available from; https://pubmed.ncbi.nlm.nih. BCN02 HIV-1 Kick&Kill clinical trial. Front Immunol. 2020;11:418. gov/17912361/. Available from; www.frontiersin.org. 50 Zimmermann MT, Oberg AL, Grill DE, et al. System-wide associa- 61 Corley MJ, Pang APS, Rasmussen TA, Tolstrup M, Søgaard OS, tions between DNA-methylation, gene expression, and humoral Ndhlovu LC. Candidate host epigenetic marks predictive for HIV immune response to influenza vaccination. PLoS One. 2016;11:(3) reservoir size, responsiveness to latency reversal, and viral rebound. e0152034. Available from; www.synapse. AIDS. 2021. Available from; http://www.ncbi.nlm.nih.gov/ 51 Sui Y, Berzofsky JA. Myeloid cell-mediated trained innate immunity in pubmed/34482353. mucosal AIDS vaccine development. Front Immunol. 2020;11:315. Fron- 62 Ruiz-Riol M, Brander C. Can we just kick-and-kill HIV: possible tiers Media S.A. Available from: /pmc/articles/PMC7058986/ . challenges posed by the epigenetically controlled interplay between 52 Palgen JL, Feraoun Y, Dzangue-Tchoupou G, et al. Optimize HIV and host immunity. Immunotherapy. 2019;11(11):931–935. prime/boost vaccine strategies: trained immunity as a new player Available from; https://www.futuremedicine.com/doi/abs/ in the game. Front Immunol. 2021;12: 612747. Frontiers Media S.A. 10.2217/imt-2019-0092. Available from: /pmc/articles/PMC7982481/ . 63 Qu K, Zaba LC, Satpathy AT, et al. Chromatin accessibility land- 53 Sanchez J, Gonçalves E, Llano A, et al. Immune profiles identifica- scape of cutaneous T cell lymphoma and dynamic response to tion by vaccinomics after MVA immunization in randomized clini- HDAC inhibitors. Cancer Cell. 2017;32(1):27–41. Available from; cal study. Front Immunol. 2020;11: 586124. Available from; https:// https://pubmed.ncbi.nlm.nih.gov/28625481/. pubmed.ncbi.nlm.nih.gov/33244316/. 64 Jiang C, Lian X, Gao C, et al. Distinct viral reservoirs in individuals 54 Gallego-Pa€uls M, Hernandez-Ferrer C, Bustamante M, et al. Variability with spontaneous control of HIV-1. Nature. 2020;585(7824):261– of multi-omics profiles in a population-based child cohort. BMC Med. 267. Available from; https://www.nature.com/articles/s41586- 2021;19(1):1–16. Available from; https://bmcmedicine.biomedcentral. 020-2651-8. com/articles/10.1186/s12916-021-02027-z. 65 Hamann MV, Ehmele P, Verdikt R, et al. Transcriptional behavior 55 Hofmeister BT, Lee K, Rohr NA, Hall DW, Schmitz RJ. Stable of the HIV-1 promoter in context of the BACH2 prominent proviral inheritance of DNA methylation allows creation of epigenotype integration gene. Virus Res. 2021;293. Available from; https:// maps and the study of epiallele inheritance patterns in the absence pubmed.ncbi.nlm.nih.gov/33316352/. of genetic variation. Genome Biol. 2017;18(1):155. Available from; 66 Cole B, Lambrechts L, Gantner P, et al. In-depth single-cell analysis https://pubmed.ncbi.nlm.nih.gov/28814343/. of translation-competent HIV-1 reservoirs identifies cellular sour- 56 Moron-Lopez S, Kim P, Søgaard OS, Tolstrup M, Wong JK, Yukl ces of plasma viremia. Nat Commun. 2021;12(1):1–13. Available SA. Characterization of the HIV-1 transcription profile after romi- from; https://www.nature.com/articles/s41467-021-24080-1. depsin administration in ART-suppressed individuals. AIDS. 67 Maldarelli F, Wu X, Su L, et al. Specific HIV integration sites are 2019;33(3):425–431. Available from; https://pubmed.ncbi.nlm.nih. linked to clonal expansion and persistence of infected cells. Science. gov/30531314/. 2014;345(6193):179. Available from: /pmc/articles/PMC4262401/ . 57 Chen HC, Martinez JP, Zorita E, Meyerhans A, Filion GJ. Position 68 Wagner TA, McLaughlin S, Garg K, et al. Proliferation of cells with effects influence HIV latency reversal. Nat Struct Mol Biol. 2017;24 HIV integrated into cancer genes contributes to persistent infec- (1):47–54. Available from; https://www.nature.com/articles/ tion. Science. 2014;345(6196):570. Available from: /pmc/articles/ nsmb.3328. PMC4230336/ . www.thelancet.com Vol 78 Month April, 2022 19