Medications for opioid use disorder shape immune responses during chronic HIV infection

Summary

People living with HIV (PLWHs) have higher risk of opioid use disorder (OUD). Whether medications for opioid use disorder (MOUDs) change immune responses in HIV infection is unknown. We examined the immune profiles in PLWHs before and 3 months after initiation of the μ opioid receptor agonist methadone, partial agonist buprenorphine, and antagonist naltrexone. Using single-cell DOGMA-seq, we profiled 29,462 peripheral blood immune cells in 12 PLWHs. We found that naltrexone treatment increased type I interferon (IFN) responses while buprenorphine increased tumor necrosis factor (TNF) responses in cytotoxic T cell population. We found that HIV+ cells in PLWHs with OUD upregulated PTPN13 and TAF5L, both of which are associated with HIV replication. We found trends suggesting increased HIV RNA expression after methadone and decreased HIV RNA expression after buprenorphine and naltrexone initiation. Overall, PLWHs treated with MOUD had improved immune responses and decreased HIV expression.

Graphical abstract

Highlights

• •Opioid μ receptor agonists do not change the size of HIV reservoir
• •T cells and dendritic cells in the blood express μ1 opioid receptor
• •μ receptor partial agonist buprenorphine and antagonist naltrexone shape immunity
• •HIV+ cells in people with opioid use upregulated PTPN13 and TAF5L

Peripheral blood immune cells, particularly T cells and dendritic cells, express detectable levels of μ opioid receptor RNA. In vivo administration of medication for opioid use disorder, namely μ opioid receptor agonist methadone, partial agonist buprenorphine, and antagonist naltrexone, shapes immune responses in people with HIV having opioid use disorder.

Introduction

The opioid epidemic has led to over 1 million deaths in the United States (US) since 1965 with the majority occurring since 1999. In 2021, 2.5 million people in the US suffered from an opioid use disorder (OUD).¹ Opioid-involved overdose deaths increased from 49,860 in 2019 to 81,806 in 2022.² Increasing opioid use has led to an increased need for medications for opioid use disorder (MOUDs) to reduce addiction and death by overdose.

Opioids bind to opioid receptors on neurons to reduce pain. However, the rewarding euphoric effects can lead to OUD. MOUDs (particularly methadone and buprenorphine) have significantly reduced rates of lethal opioid overdose.³ The opioid receptor family proteins include μ, δ, and κ opioid receptors and nociception receptor (ORL1).⁴ The μ opioid receptor is responsible for addiction and is associated with euphoria.⁵ Persons with OUD suffer withdrawal symptoms⁶ in the absence of μ opioid receptor agonism. Medications targeting μ opioid receptors can eliminate addiction symptoms, relieve drug craving, and thus treat OUD. Methadone (a μ opioid receptor agonist), buprenorphine (a μ opioid receptor partial agonist), and naltrexone (a μ opioid receptor antagonist) are the only three Food and Drug Administration (FDA)-approved forms of MOUD that effectively treat OUD through reducing opioid craving and use. Methadone is a μ opioid receptor agonist that blocks the reinforcing euphoric effects and reverses withdrawal symptoms caused by other opioids.⁷ Buprenorphine is a high-affinity and low-intrinsic-activity μ receptor partial agonist, which displaces opioid full agonists from the receptor and reduces opioid use.^8,9 Naltrexone is a μ opioid receptor antagonist and a weak δ and κ opioid receptor antagonist that blocks the reinforcing euphoric effect of opioids.¹⁰ Naloxone is sometimes co-formulated with buprenorphine in the transmucosal/sublingual tablet formulations where it does not function as an antagonist due to limited bioavailability when administered sublingually. The co-formulation of naloxone with buprenorphine was to prevent the misuse of the buprenorphine as an injected drug without actual antagonist activity when used as prescribed, as naloxone fully blocks μ receptor activity when injected.¹¹

People who inject drugs have higher incidence of HIV infection, reduced access to antiretroviral therapy (ART) and MOUD, and higher mortality.^12,13 Sharing of injection drug use equipment in prescription opioid oxymorphone misuse led to an HIV outbreak in rural Indiana,¹⁴ and injection of heroin, fentanyl, and stimulants have led to numerous other HIV outbreaks in the US. MOUD improves HIV prevention,^15,16 increases HIV viral suppression,^17,18,19 and reduces overall mortality.¹² Despite these benefits, only 1 in 5 persons with OUD receives an effective form of MOUD.²⁰ Long-term implementation of both MOUD and HIV treatment can synergistically stop the opioid epidemic, reduce HIV transmission, treat OUD, and reduce HIV-related mortality.

While opioids mainly affect pain reception in the peripheral nervous system and the reward circuits in the central nervous system, earlier studies suggest that immune cells,²¹ including monocytes and T lymphocytes, also express μ,^{22,23,24,25,26,27,28} δ,^{29,30,31,32,33,34} and κ^{34,35,36,37,38,39,40,41,42} opioid receptors.⁴³ Thus, opioids may impact immune cell function. Chronic opioid use affects HIV infection at three levels.^44,45 First, opioids induce systemic immune dysfunction by suppressing type I interferon (IFN) responses in macrophages,⁴⁶ suppressing IFNγ promoter activity⁴⁷ and production,⁴⁸ inducing aberrant interleukin-4 expression,⁴⁹ decreasing T cell activation (by inhibiting nuclear factor κB [NF-κB] activation),⁵⁰ and reducing peripheral blood lymphocyte proliferation.^35,51 Second, during acute or uncontrolled HIV infection, opioids increase expression of HIV corepressors CXCR4⁵² and CCR5,^52,53,54 increasing HIV infectivity.^55,56 Third, after HIV establishes infection and integrates into the genome of infected cells, opioids may affect HIV expression with contradicting results: opioids may diminish HIV reactivation,^55,57 increase HIV reactivation,^58,59 or have no effect.⁵⁶ Importantly, opioid receptor blockade by naloxone or naltrexone can prevent or reverse some of these opioid-induced adverse effects, such as restoring IFNγ expression,⁴⁸ maintaining lymphocyte proliferation,^60,61 and reducing opioid-induced CCR5 upregulation and increased HIV infection.^53,55 These complex and puzzling results reflect a significant knowledge gap of how opioids affect systemic immune responses and HIV infection. Understanding the impact of chronic opioid use and treatment on immune responses and HIV persistence is essential for controlling both the opioid and the HIV epidemics.

Here, in our prospective, open-label clinical trial,⁶² we recruited people living with HIV (PLWHs) under suppressive ART who also had Diagnostic and Statistical Manual (DSM)-5 moderate to severe OUD and were seeking a form of MOUD. These participants were provided infrastructural support for three participant-selected MOUDs: methadone, buprenorphine, or naltrexone. As previously published in a protocol paper, we aimed to examine the impact of the 3 FDA-approved forms of MOUD on host immune responses and the HIV reservoir, detailed in our previous publication.⁶² To examine heterogeneous immune cells and identify the rare (<0.1% in the peripheral blood CD4⁺ T cells) HIV-infected cells, we used single-cell multiomic DOGMA-seq to examine host systemic immune responses and individual HIV-infected cells at epigenetic (by assay for transposase-accessible chromatin using sequencing, ATAC-seq), transcriptional (by RNA sequencing, RNA-seq), and protein expression (by cellular indexing of transcriptomes and epitopes by sequencing, CITE-seq) levels within the same single cells. We profiled μ opioid receptor expression in different immune cells, identified the single-cell transcriptional landscape of the rare HIV-infected cells in PLWHs receiving MOUD, and identified unique immune programs in the μ receptor partial agonist buprenorphine treatment and μ receptor antagonist naltrexone treatment arms. Importantly, this is the first study, which defines the impact of μ opioid receptor agonist (methadone), partial agonist (buprenorphine), and antagonist (naltrexone) on peripheral blood immune responses and HIV-infected cells.

Results

MOUD does not change the size of the HIV reservoir

We first examined whether the size of the HIV reservoir changes before and after μ receptor activation, namely methadone and buprenorphine treatment. From 4 participants receiving methadone and 5 receiving buprenorphine, we found that the size of the HIV reservoir (as measured by quantitative viral outgrowth culture, Figure 1), was not significantly different before and 3 months after MOUD.

Figure 1.

No change in the size of the HIV reservoir before and after methadone and buprenorphine

The size of the HIV reservoir was measured by quantitative viral outgrowth culture (paired Wilcoxon rank-sum test). Each linked pair of dots represents data from the same individual. See also Table S1.

Single-cell DOGMA-seq captures the immune cell subsets and opioid receptor expression in peripheral blood immune cells

To examine the impact of MOUD on immune responses and HIV infection, we obtained paired peripheral blood (peripheral blood mononuclear cell [PBMC] and enriched memory CD4⁺ T cells) from 4 participants in the methadone arm, 4 participants in the buprenorphine arm, and 4 participants in the naltrexone arm, at treatment initiation (baseline) and 3 months after MOUD initiation (Table S1). Of note, participants in the naltrexone arm had detectable viremia despite ART and lower CD4⁺ T cell counts (median 360/μL) than participants in the methadone (median 630/μL) and buprenorphine (median 1,290/μL) arms, and the results should be interpreted with caution. We used DOGMA-seq, a single-cell multiomic method that captures the chromatin accessibility, transcriptome, and surface protein expression of 154 markers in the same single cells. The selected panel of markers spans immune cell subsets and function including markers of cell function, activation, memory status, and localization. After removing low-quality cells and doublets, a total of 29,462 cells were analyzed, including 5,906 cells from the methadone arm, 13,477 from the buprenorphine arm, and 10,079 from the naltrexone arm. Resulting cells had a median of 4,274 subnucleosomal ATAC fragments, 2,830 RNA transcripts, and 1,456 genes per cell. After batch effect correction, features from all three modalities (ATAC-seq, RNA-seq, and CITE-seq) were projected on a shared weighted nearest neighbor (WNN) uniform manifold approximation and projection (UMAP) plot (Figure 2A). There was no discernable batch effect between individuals (Figure S1A) or treatments (Figure S1B). We identified major immune cell types in the peripheral blood (Figure 2A), including 3 naive T cell clusters, 6 polarized T cell clusters (T cells - AP-1/LTB [lymphotoxin B], T cells - LTB, T cells - CX3CR1, T cells - memory, Treg/proliferating, and cytotoxic T cells), natural killer cells, 3 B cell clusters (naive B cells, memory B cells, and plasma cells), monocytes, classical dendritic cells, and plasmacytoid dendritic cells. This identification was based on their top differentially expressed genes (Figure S2A), cellular transcriptome (Figure S2B), and surface protein expression (Figure S2C). For example, cytotoxic T cells contain a mixture of CD4⁺ and CD8⁺ T cells actively expressing cytotoxic effector genes by RNA-seq. To determine if different MOUDs affected peripheral blood immune cell composition, we examined the proportion of immune cells before and after each MOUD treatment (Figure 2B). After naltrexone treatment, there was a decrease in naive T cell proportion (29.2% at baseline versus 10.1% after naltrexone) with an associated increase in cytotoxic T cell proportion (23.5% at baseline versus 36.4% after treatment) (red arrow, Figure 2B), particularly in two participants (red arrows, Figure 2C). Importantly, we also found that peripheral blood immune cells expressed detectable levels of opioid receptors, particularly μ opioid receptors in T cells (Figure 2D).

Figure 2.

Peripheral blood immune cell composition before and after methadone (opioid μ receptor agonist), buprenorphine (opioid μ receptor partial agonist), and naltrexone (opioid μ receptor antagonist)

(A) Weighted nearest neighbor (WNN) UMAP of integrated, modality-weighted, and filtered immune cells annotated by their transcriptome and surface protein expression. Data from both PBMCs and enriched memory CD4⁺ T cells were shown on the same WNN UMAP plot (n = 29,462).

(B) Composition of peripheral blood immune cells (from PBMCs, n = 9,249) before and after MOUD.

(C) Composition of immune cells (from PBMCs) before (open circle) and after (closed circle) each MOUD at the individual participant level. Each linked pair of dots represents data from the same individual.

(D) Opioid receptor (μ1, δ1, κ1, and nociceptin) expression levels in peripheral blood immune cells (from PBMCs). Red arrowheads indicate a decrease in naive cell proportion and increase in cytotoxic T cell proportion in two naltrexone-treated individuals in (B) and (C). See also Figures S1 and S2.

Buprenorphine and naltrexone shape immune responses in cytotoxic T cells

Given the pivotal role of cytotoxic T cells in HIV persistence, we examined the cellular profiles of cytotoxic T cell populations after different MOUD treatments. In cytotoxic T cells, we found 229 differentially expressed genes between methadone and buprenorphine treatment (Figure 3A), 265 differentially expressed genes between buprenorphine and naltrexone treatment (Figure 3B), and 71 differentially expressed genes between methadone and naltrexone treatment (Figure 3C). When comparing the μ receptor antagonist (naltrexone) versus agonist (methadone) treatment, we found that naltrexone treatment had increased type I IFN response gene expression (such as IFI44L and MX1). Type I IFN responses in naltrexone treatment were also elevated compared to methadone treatment in cytotoxic T cells by gene set enrichment analysis (GSEA) (Figure 3D). Comparing μ receptor partial agonist (buprenorphine) versus agonist (methadone) treatment, we found that buprenorphine treatment showed increased tumor necrosis factor (TNF) response gene expression (TNFAIP3 and ZFP36L2). TNF responses in buprenorphine treatment were also increased (compared to methadone treatment and naltrexone treatment), in cytotoxic T cells by GSEA (Figure 3E).

Figure 3.

Naltrexone treatment increases type I IFN responses and buprenorphine increases TNF responses in cytotoxic T cell population

(A–C) Differential gene expression between different MOUD after 3 months of respective MOUD treatment.

(D and E) Fast gene set enrichment analysis using fGSEA identified enrichment in IFN-induced gene modules in naltrexone treatment (D) and enrichment of TNF signaling in buprenorphine treatment (E) in cytotoxic T cell subset. ∗, adjusted p value < 0.05.

(F) Transcription factor motif activity scores before and after MOUD. ∗, Wilcoxon rank-sum test.

We further examined the differences in transcription factor activity using ChromVAR to study chromatin accessibility of immune cells before and after different MOUD treatments (Figure 3F). We found that 3 months of buprenorphine treatment significantly increased the activity of transcription factors involving T cell activation including NF-κB (RELA) and AP-1 (FOSL1:JUND and ATF3) compared to baseline and to 3 months of methadone or naltrexone treatment. Together, these results suggest an unexpected induction of T cell activation in cytotoxic T cells possibly through TNF stimulation to the NF-κB and AP-1 pathways specifically in buprenorphine treatment.

DOGMA-seq captures the rare HIV-infected cells, identifying a trend of lower HIV expression after buprenorphine and naltrexone treatment

HIV-infected cells are extremely rare (<0.1%) in virally suppressed individuals.^{63,64,65,66,67,68} In our DOGMA-seq dataset, we mapped ATAC-seq reads to HIV genome to identify HIV DNA+ cells, as previously described.⁶³ We also mapped RNA-seq reads to HIV genome to identify HIV RNA+ cells. Of note, since Tn5 in ATAC-seq can only bind to open chromatin regions, HIV proviruses that are not accessible cannot be identified using this method. Given the sparsity of the HIV genome within the human genome, ATAC-seq-based identification of HIV provirus is not sensitive. In addition, ATAC-seq does not recover sufficient HIV DNA to determine whether the HIV provirus is intact or defective. Therefore, most HIV DNA+ cells are presumably defective.⁶⁶ We identified 21 HIV DNA+ RNA− cells (Figure 4A, blue) and 17 HIV-1 RNA+ cells (Figure 4A, magenta) across different T cell subsets (Figure 4B). Because of the low cell number of HIV-infected cells identified, we pooled the HIV DNA+ RNA− cells and HIV RNA+ cells, and cells before and after MOUD treatment, as HIV-infected cells for differential gene expression analysis (Figure 4C). We found that HIV-infected cells upregulate PTPN13 and TAF5L compared with uninfected cells from the same PLWHs (Figure 4D). PTPN13 (a non-receptor tyrosine-protein phosphatase) is a marker for Th17 polarization⁶⁹ and an HIV dependency factor.⁷⁰ TAF5L is a TATA box binding protein within the P300/CBP-associated factor (PCAF) histone acetylase complex and a part of the transcription initiation complex, and the macaque homolog is upregulated in simian immunodeficiency virus infection.⁷¹ Next, by comparing normalized HIV expression levels, we found trends suggesting increased HIV RNA expression levels after methadone treatment and decreased HIV RNA expression levels after buprenorphine and naltrexone treatments (Figure 4E). Of note, because of the low number of cells examined in a small number of participants, these results need to be interpreted with caution.

Figure 4.

The cellular program of HIV-infected cells in the context of OUD

(A) Identification of HIV-infected cells in PLWHs before and after MOUD, as shown on a WNN UMAP including data from both PBMCs and memory CD4⁺ T cells. n = 21 for HIV DNA+ RNA− cells (blue) and n = 17 for transcriptionally active HIV-1 RNA+ cells (magenta).

(B) Composition of HIV-infected cells in the CD4⁺ T cell clusters in enriched memory CD4⁺ T cells (n = 17,980).

(C) Differentially expressed genes in HIV-infected versus uninfected cells.

(D) Expression level of differentially expressed genes in HIV-infected versus uninfected cells. ∗, adjusted p < 0.05 by Wilcoxon rank-sum test.

(E) Normalized HIV RNA expression levels before and after MOUD. ND, not detected.

Discussion

In this prospective, open-label, longitudinal study, we recruited study participants with HIV and OUD who were receiving one of the three FDA-approved forms of MOUD (the μ opioid receptor agonist methadone, partial agonist buprenorphine, and antagonist naltrexone) with paired samples before and after respective MOUD treatment. We aimed to examine how MOUD changes immune responses in PLWHs under suppressive ART. The availability of paired samples from the same individuals before and after MOUD allowed us to evaluate the impact of MOUD on HIV infection and immune response. By comparing longitudinally archived paired samples from the same participants, we identified the heterogeneous changes of immune cell subsets before and after MOUD (Figure 2). By comparing differences between μ opioid receptor agonist (methadone) and antagonist (naltrexone), we identified the unique effect of μ opioid receptor blockade in restoring type I IFN responses in vivo (Figure 3D). We found that the partial agonist buprenorphine treatment showed increased TNF responses in cytotoxic T cells (Figure 3E). Using ATAC-seq, we found that buprenorphine treatment increased NF-κB and AP-1 transcription factor activity compared to that of methadone and naltrexone (Figure 3F). Overall, our study demonstrates how different forms of MOUD affect peripheral immune programs in PLWHs.

HIV-infected cells are extremely rare (<1%) in PLWHs under suppressive ART.⁶⁶ Whether opioid use increases HIV reactivation remains under debate. Previous published studies have found that opioids can increase HIV reactivation,⁵⁸ decrease HIV reactivation,⁵⁷ or have no effect.⁵⁶ These studies tested HIV reactivation in cell line models or under ex vivo stimulation. Furthermore, these studies utilized bulk profiling of HIV RNA expression, such as cell-associated RNA measurement in millions of cells. This masks critical information regarding whether the HIV RNA expression readout comes from a few cells producing many HIV RNA or many cells producing few HIV RNA. In contrast, we used single-cell DOGMA-seq to capture the rare HIV-infected cells and examine HIV RNA expression level. We found that HIV-infected cells in PLWHs after OUD upregulate two cellular factors, PTPN13 and TAF5L, which are associated with HIV replication (Figure 4C). In addition, we found a trend that in vivo use of the opioid partial agonist buprenorphine or antagonist naltrexone may suppress HIV expression (Figure 4E). Whether these effects of MOUD persist with long-term treatment to reduce HIV-induced immune activation and inflammation remains to be determined.

In our study without ex vivo stimulation, we found that cytotoxic T cells upregulate type I IFN response genes after 3 months of μ opioid receptor blockade by naltrexone. In a previous study profiling how opioids affect immune responses in people with OUD,⁷² opioid use decreased type I IFN responses in lipopolysaccharide (LPS)-stimulated CD4⁺ T cells ex vivo and in LPS-stimulated CD4⁺ T cells treated with morphine (10–100 μM) in vitro. Our results suggest that opioid-mediated suppression of type I IFN responses can be restored by in vivo μ opioid receptor blockade by MOUD, providing evidence that MOUD may improve immune functions in PLWHs.

In summary, we found that μ opioid receptor blockade by naltrexone restores type I IFN responses, increases AP-1 and NF-κB activity in cytotoxic T cell population, and may decrease HIV reactivation in PLWHs.

Limitations of the study

The low number of study participants is a major limitation for this study. Although we prospectively screened 510 PLWHs having OUD, participant retention in the follow-up visits, blood sampling in people with difficult venous access, and having enough cells for multiomic profiling and viral outgrowth cultures remained challenging. The biological differences in participants, such as CD4⁺ T cell count, viral load, treatment history, and hepatitis C viral infection (HCV) status, may further limit the generalizability of this study. The study intervention was also limited at three months, which may not be sufficient time to reverse HIV-1- and OUD-related effects on peripheral blood immune cells. The dynamics and extent of the responses to MOUD can also depend on factors such as opioid doses, MOUD doses, length of opioid exposure before MOUDs, adherence to MOUDs, and the type of opioids previously used by participants. The baseline heterogeneity may affect longitudinal outcomes in terms of sensitivity to MOUD effects and the speed of immune and viral changes following MOUDs and concomitant opioid suspension. In addition, our study participants are mainly male. Sex as a biological variable may not be fairly compared in this study with limited recruitment of female participants. Finally, our measures of viral reservoir with quantitative viral outgrowth assay were limited to CD4⁺ T cells in peripheral blood, which does not capture the full diversity of infected peripheral blood cells and immune cells in tissues.

Study approval

This study was approved by Yale Institutional Review Board (IRB). All participants provided informed consent.⁶²

Resource availability

Lead contact

Requests for further information, resources, reagents, and codes should be directed to and will be fulfilled by the lead contact, Dr. Ya-Chi Ho (ya-chi.ho@yale.edu).

Materials availability

This study did not generate any new unique reagents or models.

Data and code availability

• •Data availability: single-cell ATAC-seq and RNA-seq data have been deposited at the Gene Expression Omnibus (GEO) and are publicly available. Accession numbers are listed in the key resources table.
• •Code availability: all code necessary to reproduce this work has been made available at https://github.com/JackCollora/2023MorphineDOGMA.
• •All other items: any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank all participants. This work was supported by National Institutes of Health (NIH) R61/R33 DA047037 (S.A.S., Y.K., and Y.-C.H.), NIH R01 AI174863, R01 AI183430, R01 AI141009, P01 AI169768, BEAT-HIV Martin Delaney Collaboratory UM1 AI164570, REACH UM1 AI164565, CHEETAH U54 AI170856, R01 AI176601, UM1 Y-SCORCH DA051410, U01 M-SCORCH DA053628, R01 DA051906 (Y.-C.H.), DP1 DA056106 (S.A.S), and T32 AI055403 (J.A.C.).

Author contributions

Y.-C.H., S.A.S., and Y.K. conceived the project. J.A.C. and Y.-C.H. designed the experiments. S.A.S. designed the clinical trial and recruited study participants. T.A.D., S.F.S., S.Z., and R.K. processed clinical samples and archived clinical demographics. J.A.C., T.A.D., D.C.L., and A.E. performed DOGMA-seq experiments. Y.-C.H., S.A.S., and Y.K. obtained funding resources. J.A.C. and Y.-C.H. wrote the manuscript with input from all authors.

Declaration of interests

Y.-C.H. served on the advisory board of this journal. S.A.S. provided paid scientific consultation to Alkermees Inc. and received in-kind study drug donations for NIH-sponsored research from Alkermes Inc. (Vivitrol) and from Indivior Pharmaceutical Company (Sublocade).

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
TotalSeq-A Human Universal Cocktail V1.0	BioLegend	CAT # 399907; RRID: AB_2888692
TotalSeq-A anti-human Hashtag 1	BioLegend	CAT # 394601; RRID: AB_2750015
TotalSeq-A anti-human Hashtag 2	BioLegend	CAT # 394603; RRID: AB_2750016
TotalSeq-A anti-human Hashtag 3	BioLegend	CAT # 394605; RRID: AB_2750017
TotalSeq-A anti-human Hashtag 4	BioLegend	CAT # 394637; RRID: AB_2814350
TotalSeq-A anti-human Hashtag 5	BioLegend	CAT # 394609; RRID: AB_2750019
TotalSeq-A anti-human Hashtag 6	BioLegend	CAT # 394611; RRID: AB_2750020
TotalSeq-A anti-human Hashtag 7	BioLegend	CAT # 394613; RRID: AB_2750021
TotalSeq-A anti-human Hashtag 8	BioLegend	CAT # 394615; RRID: AB_2750022
Biological samples
Blood samples (Table S1)	This study
Chemicals, peptides, and recombinant proteins
Digitonin 5%	Thermo Fisher	CAT # BN2006
Human TruStain FcX	BioLegend	CAT # 422302; RRID: AB_2818986
Ficoll-Paque PLUS	Cytiva	CAT #17-1440-03
Critical commercial assays
Chromium Next GEM Single Cell Multiome ATAC + Gene Expression Reagent Bundle	10× Genomics	PN-1000283
Chromium Nuclei Isolation Kit with RNase Inhibitor	10× Genomics	PN-1000494
Chromium Next GEM Chip J Single Cell	10× Genomics	PN-1000230
Dual Index Kit TT Set A	10× Genomics	PN-1000215
Single Index Kit N Set A	10× Genomics	PN-1000212
3ʹ Feature Barcode Kit	10× Genomics	PN-1000262
EasySeq Dead Cell Removal Annexin V kit	STEMCELL	CAT # 17899
EasySep Human CD4+ T cell Isolation Kit	STEMCELL	CAT # 17952
Memory CD4+ T cell Isolation Kit	Miltenyi Biotec	CAT # 130-091-893
Dynabead Human T-Activator CD3/CD28 for T cell Expansion and Activation	Thermo Fisher	CAT# 11131D
HIV p24 ELISA	Perkin Elmer	NEK050B
Deposited data
DOGMA-seq dataset	This study	GSE234521
Code used in this study	This study	https://github.com/JackCollora/2023MorphineDOGMA
Software and algorithms
cellranger-arc (version 2.0.2)	10× Genomics	https://support.10xgenomics.com/single-cell-multiome-atac-gex/software/pipelines/latest/installation#download
ArchR (version 1.0.1)	Granja et al.73	https://www.archrproject.com/
MACS2 (version 2.2.7.1)	Zhang et al.74	https://hbctraining.github.io/Intro-to-ChIPseq/lessons/05_peak_calling_macs.html
Kallisto (version 0.48.0)	Bray et al.75	https://pachterlab.github.io/kallisto/
Bowtie2 (version 2.4.2)	Langmead et al.76	https://github.com/BenLangmead/bowtie2
STAR (version 2.7.11a	Dobin et al.77	https://github.com/alexdobin/STAR
Seurat (version 4.3.0)	Hao et al. and Hao et al.78,79	https://cran.r-project.org/web/packages/Seurat/index.html
Signac (version 1.9.0)	Stuart et al.80	https://cran.r-project.org/web/packages/Signac/index.html
fastMNN (version 1.14.1)	Haghverdi et al.81	https://rdrr.io/github/LTLA/batchelor/man/fastMNN.html
Harmony (version 0.1.1)	Korsunsky et al.82	https://portals.broadinstitute.org/harmony/articles/quickstart.html
chromVAR (version 1.20.2)	Schep et al.83	http://bioconductor.org/packages/release/bioc/html/chromVAR.html
JASPAR2022 R package (version 0.99.7)	Castro-Mondragon et al.84	https://bioconductor.org/packages/release/data/annotation/html/JASPAR2022.html
fGSEA (version 1.24.0)	Korotkevich et al.85	https://bioconductor.org/packages/release/bioc/html/fgsea.html
MSigDB (version 7.5, sourced from R package msigdbr version 7.5.1)	Dolgalev et al.86	https://cran.r-project.org/web/packages/msigdbr/vignettes/msigdbr-intro.html
R (version 4.2.0)	R Core Team	https://www.r-project.org/

Experimental model and study participant details

Study participants

The recruitment of study participants was described previously.⁶² Briefly, in our prospective, open-label study,⁶² we proactively screened 510 adult PLWH individuals under suppressive ART (plasma viral load <200 copies/mL) who also have moderate to severe OUD (defined by DSM-5 criteria). Due to low enrollment in the naltrexone arm, PLWH (under suppressive ART) who were receiving naltrexone for alcohol use disorder with a history of OUD were also enrolled. Over the course of 3.5 years (encompassing COVID restriction in research activities), 27 eligible participants were successfully enrolled for peripheral blood sampling before (baseline) and after 3 months of MOUD. Sample size (samples from 4 donors in each MOUD arm) was determined by sample availability and considerations for a balance of sex, race, and age. Participants opted MOUD based on physician recommendations and their own decisions.

Method details

Frequency of HIV-infected cells

The size of the HIV reservoir as measured by quantitative viral outgrowth assay as previously described.^65,66 Briefly, CD4⁺ T cells from PLWH were isolated from PLWH by negative depletion (EasySep Human CD4⁺ T cell Isolation Kit, STEMCELL) and plated at limiting dilution and stimulated with CD3/CD28 (Dynabeads Human T-Activator CD3/CD28, Thermo Fisher) and IL-2. Three days after activation, MOLT4-CCR5 cell line were added to the culture wells and cultured for 3 weeks, with media changes twice a week. The presence of replication competent HIV was measured by p24 ELISA of the culture supernatant. The frequency of HIV-infected cells was calculated using IUPMStats.⁸⁷

DOGMA-seq

Peripheral blood mononuclear cells (PBMC) were isolated by Ficoll density gradient centrifugation (Ficoll-Paque PLUS, Cytiva #17-1440-03) and cryopreserved. From cryopreserved PBMC from each participant, dead cells were removed by magnetic depletion (EasySep Dead Cell Removal (Annexin V) Kit, STEMCELL Technologies, #17899). PBMC and memory CD4⁺ T cells (enriched using Miltenyi human memory CD4⁺ T cell isolation kit, #130-091-893) from each participant were each stained with Totalseq-A hashtag antibodies (BioLegend) and pooled for DOGMA-seq,⁸⁸ as we previously described.^63,64,89 Briefly, aliquots of ∼0.5 million PBMC per sample and ∼1 million memory CD4⁺ T cells were stained with barcoded TotalSeq-A hashing antibodies (BioLegend), incubated at 4°C for 30 min, and washed for 3 times. Equal numbers of PBMC (∼62,500 PBMC/sample, each stained with a unique hashtag, 8 samples per pool) were pooled into 0.5 million cells. Equal numbers of memory CD4⁺ C cells (∼125,000 memory CD4⁺ T cells/sample, each stained with a unique hashtag, 8 samples per pool) were pooled into 1 million cells. Care was taken to ensure each pool was a mixture of treatment conditions and participants to avoid confounding batch with a biological factor of interest.

The pooled samples were stained with a panel of 154 cell surface proteins and 9 isotype controls included in TotalSeq-A Human Universal Cocktail (V1.0, BioLegend, catalog # 399907). All antibodies were prepared following manufacturer protocols. For each 0.5 million of the pooled samples, cells were resuspended in 22.5 μL staining buffer and incubated with 2.5 μL Human TruStain FcX (BioLegend, catalog # 422302) at 4°C for 10 min. Cells were then incubated with 25 μL of the TotalSeq-A Human Universal Cocktail at 4°C for 30 min and washed for 3 times.

Cells were then permeabilized according to the DIG lysis protocol from the original DOGMA-seq protocol (Mimitou et al., 2021). Briefly, digitonin lysis buffer (0.01% digitonin, 20 mM Tris-HCl pH 7.4, 150 mM NaCl, 3 mM MgCl₂, and 2 U/μL RNase inhibitor (Sigma-Aldrich, catalog # 3335402001)) and digitonin wash buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 3 mM MgCl₂, and 1 U/μL RNAse inhibitor) were freshly prepared and chilled on ice. Antibody-stained cells were resuspended in 100 μL digitonin lysis buffer on ice for 5 min. 1 mL of digitonin wash buffer was added and cells were pelleted (500 g for 5 min at 4°C). Cells were then resuspended in 7 μL digitonin wash buffer and counted using Trypan blue staining to verify permeabilization (>90%) and nuclei number.

Cells were processed according to the Chromium Next GEM Single Cell Multiome ATAC + Gene Expression protocol (10× documentation GC000338 Rev A) with modifications described in the original DOGMA-seq protocol. Libraries were sequenced on NovaSeq 6000.

Bioinformatic analysis

Alignment of RNA-seq and ATAC-seq reads to human transcriptome

RNA-seq and ATAC-seq reads were mapped to hg38 using cellranger-arc (version 2.0.2) from 10× Genomics and the corresponding precompiled hg38 reference (version 2.0.0).

ATAC-seq peak calling and quantification

Cellranger fragment files were imported into ArchR (version 1.0.1⁷³) filtering cells to include only those with a min TSS enrichment of 2 and at least 1000 ATAC fragments. Peaks were called using MACS2 (version 2.2.7.1⁷⁴) on subnucleosomal (less than 147 bp) fragments and quantified using ArchR.⁹⁰

Cell surface marker expression quantification

Cell surface proteins were quantified by pseudoalignment using Kallisto (version 0.48.0⁷⁵) on a kite (https://github.com/pachterlab/kite) derived feature barcode reference.

Alignment of ATAC-seq and RNA-seq reads to HIV viral sequences

ATAC-seq reads were mapped to HXB2 and participant derived sequences using Bowtie2 (version 2.4.2⁷⁶) with the “--very-sensitive” preset. All ATAC-seq mapping reads were retained. The two mapping iterations were collapsed to identify only the unique reads mapping to HIV. Corresponding barcodes were extracted and used to annotate HIV DNA+ cells that were HIV RNA–. Transcriptome reads were mapped using STAR (version 2.7.11a⁷⁷) separately to HXB2, a compendium of Clade B viral sequences, or study participant-derived sequences, discarding only reads that mapped >1000 independent positions for HXB2 and participant-derived sequences and >10000 independent positions for the compendium of Clade B viral sequences. The three mapping iterations were collapsed to identify only the unique reads mapping to HIV. Corresponding barcodes were extracted and used to annotate HIV RNA+ cells.

Quality control and filtering

ATAC-seq, RNA-seq, and cell surface protein expression matrices were read into Seurat/Signac objects (Seurat version 4.3.0,^78,79 Signac version 1.9.0⁸⁰), one per sample. Singlets were identified based on hashtag expression using the Seurat function HTOdemux. Doublets were discarded, singlets were further filtered to remove those with fewer than 200 genes or greater than 20% of gene expression reads mapping to mitochondrial genes.

After removal of low quality cells, single cell multiomic datasets were integrated to remove batch effects using FastMNN⁸¹ for RNA-seq and Harmony⁸² for ATAC-seq and CITE-seq. Integrated single-cell ATAC-seq, RNA-seq, and CITE-seq results were projected into a joint (Uniform Manifold Approximation and Projection) UMAP plot using the weighted nearest neighbor (WNN).⁷⁸

Clustering, dimensionality reduction, and data integration

RNA data were log normalized in each dataset and top variable features were identified. The top 2000 features were selected that were variable in the most independent samples, with ties broken by median rank across datasets. Top features were used for integration using fastMNN (version 1.14.1⁸¹). Cell surface marker expression was normalized using Centered log-ratio, variable features were identified, and Harmony integration applied across libraries (version 0.1.1⁸²). Top ATAC-seq peaks were identified, scaled using TF-IDF, and reduced using SVD. Resulting reduced data was integrated across libraries using Harmony.⁸² The first 30 principal components of RNA, the first 18 principal components of surface protein expression, and the 2^nd through 30^th SVD components of ATAC-seq peaks were used to identify multimodal neighbors using Seurat. Multimodal neighbors were then used to perform UMAP with default Seurat parameters. Clusters were identified from the weighted nearest neighbor graph, using the SLM algorithm, with a resolution of 0.8.

Cluster annotation

Clusters were annotated iteratively. First, clusters were labeled according to canonical RNA marker genes and those defined in previous publications.^63,64,91 Second, those clusters were refined by surface marker expression based on definitions from previous publications.^63,64,92

Differential analysis

Differential analysis was performed using the Wilcox rank-sum test implemented in Seurat. For chromVAR differential analysis, the default cutoffs above were used, but the mean function used was ‘rowMeans’ to as these values were Z scores. For gene expression and surface marker expression comparisons, cutoffs of 0.25 absolute log2 fold change and a minimum of 10% of cells expressing (the default values) were used to filter genes prior to differential testing. For GSEA ranking, no genes were filtered.

Transcription factor activity analysis

Transcription factor activities were calculated using chromVAR⁸³ (version 1.20.2) using motifs from the JASPAR2022⁸⁴ R package (version 0.99.7). All plots depict bias corrected “deviation” Z scores.

Gene set enrichment analysis (GSEA)

Gene Set Enrichment Analysis (GSEA) was performed using fGSEA⁸⁵ (version 1.24.0) on all measured genes ranked by their fold change between comparison groups. Gene sets assayed are from MSigDB⁸⁶ (version 7.5, sourced from R package msigdbr version 7.5.1). All gene sets in the categories H (Hallmark), C2 (Curated), and C7 (Immune signatures) were evaluated.

Quantification and statistical analysis

All statistics were performed in R (version 4.2.0). All reported p values have been corrected for multiple hypothesis testing. Plotting was performed using ggplot2 (version 3.4.2) and associate packages. The size of the HIV reservoir measured by quantitative viral outgrowth culture was statistically tested as paired Wilcoxon rank-sum test. In differential gene expression analyses, asterisks adjusted p value < 0.05 by Wilcoxon rank-sum test.

Published: June 2, 2025