Persistent T Cell Repertoire Perturbation and T Cell Activation in HIV After Long Term Treatment

Abstract

Objective: In people living with HIV (PLHIV), we sought to test the hypothesis that long term anti-retroviral therapy restores the normal T cell repertoire, and investigate the functional relationship of residual repertoire abnormalities to persistent immune system dysregulation.

Methods: We conducted a case-control study in PLHIV and HIV-negative volunteers, of circulating T cell receptor repertoires and whole blood transcriptomes by RNA sequencing, complemented by metadata from routinely collected health care records.

Results: T cell receptor sequencing revealed persistent abnormalities in the clonal T cell repertoire of PLHIV, characterized by reduced repertoire diversity and oligoclonal T cell expansion correlated with elevated CD8 T cell counts. We found no evidence that these expansions were driven by cytomegalovirus or another common antigen. Increased frequency of long CDR3 sequences and reduced frequency of public sequences among the expanded clones implicated abnormal thymic selection as a contributing factor. These abnormalities in the repertoire correlated with systems level evidence of persistent T cell activation in genome-wide blood transcriptomes.

Conclusions: The diversity of T cell receptor repertoires in PLHIV on long term anti-retroviral therapy remains significantly depleted, and skewed by idiosyncratic clones, partly attributable to altered thymic output and associated with T cell mediated chronic immune activation. Further investigation of thymic function and the antigenic drivers of T cell clonal selection in PLHIV are critical to efforts to fully re-establish normal immune function.

Keywords: T cell activation; T cell repertoire; antiretroviral therapy; blood transcriptome; chronic inflammation; human immunodeficiency virus; people living with HIV.

Figure 1

T cell receptor alpha and beta chain sequence repertoires remain disturbed in people living with HIV (PLHIV) despite long term effective anti-retroviral therapy. (A) Total number of CDR3 sequences. (B) Repertoire richness, measured as number of unique CDR3 sequences, normalized by number of total sequences. (C) Repertoire inequality, measured as Gini index of CDR3 sequence abundance distributions. (D) Repertoire diversity, measured as Shannon entropy. PLHIV, people living with HIV; HC, Healthy controls, p values for significant differences shown for Mann-Whitney U tests. Violin plots show distribution of data with individual data points, median and interquartile range. Alpha chain sequences are shown on the left and beta chain sequences on the right.

Figure 2

People living with HIV (PLHIV) show increased oligoclonal expansion of CDR3 sequences. (A) Frequency distribution of all CDR3 sequences, showing the proportion of the total repertoire that is occupied by each 10 percentile range of unique sequences. The percentile ranges are shown in decreasing order from the most abundant 10% (red) to the least abundant 10% (yellow at the top of each bar stack). Bars=median, error bars=interquartile range. (B) Frequency distribution of the 100 most abundant CDR3 sequences. (C) The mean abundance of the 100 most abundant CDR3 sequences in each individual.

Figure 3

People living with HIV (PLHIV) do not show significant enrichment of HIV, CMV and EBV-reactive CDR3 sequences. Frequency of (A) HIV, (B) CMV and (C) EBV–reactive sequences, identified through public annotation on the VDJdb database, as a proportion of the total number of the 100 most abundant alpha and beta chain CDR3 sequences. PLHIV, people living with HIV; HC, Healthy controls; p values for significant differences shown for Mann Whitney U tests. Violin plots show distribution of data with individual data points, median and interquartile range.

Figure 4

Reduced intra-individual similarity of CDR3 sequences in people living with HIV (PLHIV). (A) Network graphs showing clusters of related alpha or beta chain CDR3 sequences from representative repertoires of healthy controls (HC) or PLHIV. Networks were created from the 2500 most abundant CDR3 sequences in each repertoire. Each node represents a unique CDR3 sequence, with the node diameter proportional to its abundance in the repertoire. Two CDR3 nodes are connected by an edge if they differ from each other by a Levenshtein distance of one. Only clusters with four or more nodes are shown. (B) Number of alpha or beta chain CDR3 clusters in each individual. (C) Median cluster size (number of nodes) for alpha or beta chain CDR3 clusters in each individual. p values for significant differences shown for Mann Whitney U tests. Violin plots show distribution of data with individual data points, median and interquartile range.

Figure 5

A subset of people living with HIV (PLHIV) have more T cell receptors (TCRs) with unusually long CDR3 sequences. (A) Frequency distributions of Levenshtein distances among alpha and beta CDR3 amino acid sequences, integrated across all PLHIV or all healthy controls (HC). Pairwise Levenshtein distances were calculated for the 2500 most abundant CDR3s in each repertoire, or restricted to CDR3s that were present at least three times (the 980 most abundant sequences in each repertoire) or only once (the 980 least abundant sequences in each repertoire). One beta sample from the PLHIV group was excluded from the analysis of the top 980 sequences as it contained <100 CDR3s with abundance ≥3. (B) Heatmaps of Levenshtein distance distributions among the least abundant sequences as defined in (A). Each column represents the repertoire from an individual subject. (C) Frequency distributions of CDR3 amino acid sequence lengths among alpha and beta chains, integrated across all PLHIV or HC. Length distributions were determined for the whole repertoire, or restricted to a subset of CDR3 sequences as defined in (A). (D) Heatmaps of CDR3 length distributions among the least abundant sequences.

Figure 6

Reduced inter-individual sharing of identical CDR3 sequences in people living with HIV (PLHIV). (A) The proportion of identical CDR3 sequences that are shared between repertoires of each pair of healthy controls (HC) or PLHIV, calculated as Jaccard index. p values for significant differences shown for Mann Whitney U tests. Violin plots show distribution of data with median and interquartile range. (B) Frequency distributions showing the number of identical alpha or beta CDR3 sequences that are found in the repertoires of 6 to 12 out of 12 HC (red) or PLHIV (blue). To correct for the larger size of the PLHIV cohort (n=26 versus n=10 HC), the plots show the average results (mean+SD) of analyzing 100 random samples of 12 out of the 26 PLHIV samples available.

Figure 7

Global immune activation at steady state in people living with HIV (PLHIV) despite effective anti-retroviral therapy. (A) Modified molecular degree of perturbation (MDP) of healthy controls (HC) and PLHIV. p values for significant differences shown for Mann Whitney U tests. Violin plots show distribution of data with individual data points, median and interquartile range. (B) Volcano plot showing statistical significance against quantitative gene expression differences between PLHIV and HC. The red dashed line indicates a false discovery rate (FDR) of 0.05, equivalent to -log₁₀ FDR of 1.3. Genes highlighted in red are considered differentially expressed (n=353; n=281 higher in PLHIV and n=72 higher in HC). (C) Network diagram showing predicted upstream regulators of interacting differentially expression genes (see Supplementary Figure 5 ), found to be enriched in PLHIV. Blue nodes represent the ten most significant upstream regulators, with label size proportional to the -log₁₀ enrichment p value. Red nodes represent the subset of the 149 interacting genes that are downstream of these regulators. (D) Spearman rank correlation of TCR diversity (Shannon Index) and mean expression of target genes (red nodes) in (C), among all participants.

Figure 8

Anti-retroviral therapy (ART) attenuates or resolves the profound impact of chronic HIV infection on transcriptional signatures. (A) Mean expression of a gene signature (module) associated with untreated HIV derived from microarray data, in blood from a previous cohort of HIV-negative patients, and HIV-positive patients before and 3 months after ART (left panel), and from RNAseq data in blood from healthy controls (HC) and PLHIV described in the present study (right panel). (B) Mean expression of a type-1 interferon (IFN) inducible gene signature (module) derived from monocyte derived macrophages, in blood from a previous cohort of HIV-negative patients, and HIV-positive patients before and 3 months after ART (left panel), and from RNAseq data in blood from healthy controls (HC) and PLHIV described in the present study (right panel). (C) Mean expression of a type-1 interferon (IFN) inducible gene signature (module) derived from T cells, in blood from a previous cohort of HIV-negative patients, and HIV-positive patients before and 3 months after ART (left panel), and from RNAseq data in blood from healthy controls (HC) and PLHIV described in the present study (right panel). p values for significant differences shown for Mann Whitney U tests. Violin plots show distribution of data with individual data points, median and interquartile range.