PD-1 Expression and Cytokine Secretion Profiles of Mycobacterium tuberculosis-Specific CD4+ T-Cell Subsets; Potential Correlates of Containment in HIV-TB Co-Infection

HIV co-infection is an important risk factor for tuberculosis (TB) providing a powerful model in which to dissect out defective, protective and dysfunctional Mycobacterium tuberculosis (MTB)-specific immune responses. To identify the changes induced by HIV co-infection we compared MTB-specific CD4+ responses in subjects with active TB and latent TB infection (LTBI), with and without HIV co-infection. CD4+ T-cell subsets producing interferon-gamma (IFN-γ), interleukin-2 (IL-2) and tumour necrosis factor-alpha (TNF-α) and expressing CD279 (PD-1) were measured using polychromatic flow-cytometry. HIV-TB co-infection was consistently and independently associated with a reduced frequency of CD4+ IFN-γ and IL-2-dual secreting T-cells and the proportion correlated inversely with HIV viral load (VL). The impact of HIV co-infection on this key MTB-specific T-cell subset identifies them as a potential correlate of mycobacterial immune containment. The percentage of MTB-specific IFN-γ-secreting T-cell subsets that expressed PD-1 was increased in active TB with HIV co-infection and correlated with VL. This identifies a novel correlate of dysregulated immunity to MTB, which may in part explain the paucity of inflammatory response in the face of mycobacterial dissemination that characterizes active TB with HIV co-infection.


Introduction
HIV co-infection is the greatest risk factor for progression of latent tuberculosis infection (LTBI) to active tuberculosis (TB) [1] and is associated with increased risk of de novo infection and reinfection [2]. Data from SIV or HIV-TB co-infection indicate an increase in the risk of disease reactivation prior to the onset of severe immunosuppression [3,4]. This risk is not eliminated with HIV therapy remaining at least twice that of the general population [5]. HIV co-infection impacts the clinical phenotype of TB resulting in reduced cavitation, more severe disease, increased bacillary dissemination and poorer outcomes in terms of morbidity and mortality.
This early, specific and lasting effect on MTB-specific immunity makes HIV and LTBI coinfection a good model in which to dissect out defective, protective and dysfunctional MTBspecific immune responses. LTBI may be viewed as the successful containment of infection by effective anti-MTB immunity. Conversely, at the opposite end of the spectrum, active TB in the setting of HIV co-infection represents a manifest failure of MTB resistance control characterised by bacillary dissemination, high bacterial load and higher mortality. Cellular immune responses that specifically characterise this state, including cellular senescence [6] or immune exhaustion phenotypes, may provide insight into cellular mechanisms of failed MTB-specific immunity. MTB-specific cell subsets that are preferentially impaired or depleted by HIV infection are therefore candidates for correlation of immune containment and protective immunity.
Programmed death-1 (PD-1/CD279) is expressed by monocytes, CD4, CD8, NKT and B cells and binds to its ligands PDL-1 and 2 inhibiting proliferation and cytokine production [7]. Expression on total CD4+ and CD8+ T-cells correlates with viral load (VL) and on HIV-specific CD8+ T-cells mediates immune dysfunction. Blockade of the PD-1 axis on HIV-specific CD4+ and CD8+ cells can reverse this phenotype [7][8][9][10][11]. The PD-1 knock out mouse demonstrated a highly inflammatory response to MTB infection [12]. In humans with active TB, PD-1 expression correlated with IFN-γ production and prevention of ligand binding enhanced cellular immune function [13]. The effect of HIV and MTB co-infection in active and LTBI on the PD-1 axis in MTB-specific T-cells has not hitherto been investigated.
We hypothesized that in the setting of MTB infection, HIV co-infection would be associated with a reduced frequency of particular CD4+ MTB-specific T-cell cytokine secreting subsets and that this might serve as a correlate of MTB infection containment. Secondly, we hypothesised that HIV-TB co-infection would be associated with increased expression of PD-1 indicating a possible cellular correlate for the deficit in MTB-specific immunity. We studied four key clinical patient subgroups to delineate the impact of HIV in both stages of MTB infection. We used multi-parameter flow cytometry to measure three canonical cytokines and PD-1 expression in patients with active TB, TB and HIV co-infection, LTBI and HIV co-infection and LTBI as well as markers of T-cell exhaustion. This enabled simultaneous definition of functional and phenotypic MTB-specific T-cell profiles at the single-cell level.

Definition of LTBI
A positive IFN-y response to ESAT-6 and CFP-10 and/or positive response to novel validated MTB-specific peptides including MTB-specific peptides from Rv3873 [14,15] and Rv3615c [16]. Individuals with HIV co-infection were enrolled if they had risk factors for LTBI and screened for the presence of LTBI [17]. All individuals with LTBI had a positive interferon gamma release assay (IGRA) for MTB infection and/or tuberculin skin test.

Antigens
PPD was obtained from Serum Statens Institute and re-suspended in RPMI at 16.7μg/ml final concentration. Pools of MTB-specific 15-mer overlapping peptides covering each of ESAT-6, CFP-10, and Rv3615c were re-suspended in a single mixture (10μg/ml per peptide final concentration).

IFN-γ MTB ELISpot
Fresh or frozen peripheral blood mononuclear cells (PBMCs), 2.5 x 10 5 per well, were left unstimulated or stimulated overnight (37°C 5% CO 2 for 16 to 20 hours) with PHA, purified protein derivative (PPD) and pools of MTB-specific 15-mer overlapping peptides covering each of ESAT-6, CFP-10, EspC, TB7.7, Rv3879c, Rv3873 and Rv3878 in an IFN-γ capture ELI-Spot assay as previously described [18]. Spot forming cells were counted using an AID ELISpot reader. A positive response was at least 5 spot-forming cells more than the negative control well and at least twice the negative control well. This method of diagnosing MTB infection in the context of clinical data has been widely validated [19][20][21][22].

Intracellular cytokine staining
The cells were prepared as previously described [23], thawed PBMCs (3-5 x 10 6 per well) were cultured for 16 hours (37°C 5% CO 2 ) in 10% human serum in Roswell Park Memorial Institute (RPMI) medium. Cells were stimulated with PMA-Ionomycin, (5ng/ml PMA and 500ng/ml ionomycin final concentration), PPD or MTB-specific peptides or left unstimulated. Where possible, 1x10 6 events were acquired using a BD LSR-II flow cytometer and compensation parameters set with Anti-Rat and Anti-Mouse Ig compensation beads (BD Biosciences). Fluorescence minus one (FMO) controls were included each time. Data on CD8, CCR7, CD45RA and CD127 are not included in this report, having been published separately but are included in the methods for accuracy [23,24].

Data analysis
FlowJo©TreeStar, Inc. v9.2 was used to gate on live cells, singlets and lymphocytes and CD3 +CD4+ subsets. FMO controls were used to define positive gates for IFN-γ, IL-2 and TNF-α responses and PD-1 expression. Non-overlapping subsets of cytokine response were created using Boolean analysis. Frequencies were normalised to the unstimulated control. Responders had a response 2 times the background and >0.001% of CD3+CD4+ T-cells and were analysed for PD-1 expression. Those without a positive response were excluded from phenotypic analysis. Co-stimulatory molecules were not included in the assay to avoid signals not necessarily present in vivo. Using these stringent cut-off criteria, antigen-specific rare events were assumed to be real events.

Statistical analysis
This was conducted using IBM SPSS Statistics version 20 and Prism version 5. The Mann-Whitney U test was used for two sample comparisons and Kruskal-Wallis test with Dunn's post-test comparison for multiple comparisons. Spearman's Rank coefficient for tests of correlation of non-parametric data was used. CD4 count and HIV VL were analysed as continuous variables. A VL of 10 was designated where viral suppression was achieved.

Participant characteristics
All individuals had a positive response to ESAT-6 or CFP-10 in the MTB-specific IFN-γ ELI-Spot or T.SpotTB except one who had a TST >15mm and a positive response to Rv3615c. Routine HIV-screening is currently not undertaken in the UK for LTBI.

CD4+ MTB-specific cells secreting IFN-γ and IL-2 may serve as a correlate of containment in HIV-TB co-infection
Our data indicate potential HIV-induced correlates underlying the loss of containment of MTB and the immune dysfunction in active TB contributing to the early and lasting effects of HIV co-infection and increased TB-related morbidity and mortality [4,25,26]. We previously showed that HIV infection was associated with a reduced frequency of MTB-specific IFN-γ and IL-2 dual-secreting CD4+ cells in patients with LTBI, most of whom had normal or near normal CD4 counts due to immune restoration with HAART [24]. Here we see a greater impact on this subset in patients that have not yet started HAART and furthermore an inverse correlation with HIV VL. We found very little evidence of relationships between other MTBspecific CD4+ T-cell subsets and HIV infection pinpointing IFN-γ and IL-2-dual secreting cells as uniquely impacted by HIV infection. This may reflect an increasing inability to contain MTB in individuals with uncontrolled HIV replication, although we were unable to prove causality in this cross sectional study. Our findings are consistent with the cumulative evidence for an association of antigen-specific CD4+ cells secreting IL-2 with or without IFN-γ in the containment or resolution of viral and TB infections [24,[27][28][29][30][31][32][33]. The association of HIV VL with reduced frequency of IL-2 secreting cells has been investigated but without an HIV-uninfected control group for comparison [32,34]. The relationship with increasing VL is consistent with an effect mediated by on-going viral replication, such as specific killing but teasing out the

MTB-specific CD4+ functional effector responses and PD-1 expression may reflect mycobacterial burden
The associations of MTB-specific CD4+ cells secreting IFN-γ with or without TNF-α (CD4 + effector-like cells) with CD4 count and VL that we demonstrated likely reflects increasing mycobacterial burden in patients with HIV and TB as previously shown [23,37]. Active TB with advanced HIV co-infection was characterised by a high frequency of the MTB-specific IFN-γ-only-secreting subset and the expression of PD-1 may be linked to IFN-γ secretion [13]. Increased expression of PD-1 on CD4+ MTB-specific IFN-γ-secreting T-cells was associated with HIV co-infection and correlated with VL. Given the role of PD-1 in immune regulation and exhaustion and its close correlation with total CD4+ T-cell activation and immunosuppression in HIV infection [8,11,[38][39][40], these data suggest a novel correlate of dysregulated immunity that is a feature of active TB/HIV. Whilst these changes were also noted in those with LTBI and HIV co-infection the greatest effect was in those with active TB and HIV coinfection. Although our analysis of PD-1 expression on MTB-specific T-cells showed a correlate with HIV VL this was most marked in the patients where mycobacteria were likely to be disseminated. HIV VL may therefore be acting as a surrogate marker of mycobacterial burden driving PD-1 expression through T-cell receptor stimulation.
Studies in mice and humans indicate that in TB, PD-1 might function as a self-regulatory and tissue-protective off-switch [12,13,[41][42][43][44]. Our findings on MTB-specific T-cells in HIV-TB co-infection might therefore underlie two observations that relate to the immune 'exhaustion' and tissue-protective functions of PD-1. The increased expression of PD-1 on IFN-γ-secreting CD4+ T-cells might contribute to reduced ability to control MTB infection and underpin the paucity of immunopathology as manifested by, for example, the lack of cavitatory disease resulting in the increased bacillary load and dissemination that characterize active TB/HIV.

Role of tri-functional cells
The frequency of MTB-specific CD4+ tri-functional T-cells secreting IFN-γ, IL-2 and TNF-α was independent of both mycobacterial burden and HIV co-infection. Tri-functionality has been associated with active TB [45,46] but data from the mouse lung suggests an alternative protective role [47] and polyfunctionality is a putative marker of protective immunity in vaccine trials [48]. We did not find sufficient evidence to suggest the tri-functional CD4+ T-cell subset correlates with protection although there was a weak association in those who were highly immunosuppressed. Some data have suggested that polyfunctional, less differentiated cells might be increased at the site of extra-pulmonary TB in those with HIV co-infection [49]. Our data indicated that in PBMCs, neither the presence of polyfunctionality nor the frequency of tri-functional cells was significantly affected by HIV co-infection or TB disease stage, which detracts from its putative role as a peripheral marker of protection [50] or active TB [46], however in our limited sample size the presence of weak associations cannot be excluded.

Effect of HAART on MTB-specific CD4+ cell responses
The natural history of HIV and TB co-infection means that extricating the effects of HAART from the effects of mycobacterial burden is challenging and beyond the scope of this study. We were unable to control for the presence of HAART and although HIV VL may be taken as a surrogate marker for its effects, the immunological outcomes we observed cannot be disentangled from associated changes in mycobacterial burden.

Limitations of this work
Comprehensive analysis of a small number of precisely defined subjects has enabled us to study MTB-specific T-cell responses in detail and to uncover the most significant relationships. Other possible immunological effects cannot be ruled-out on account of our small sample size. We are unable to infer causality based solely on the associations identified in this cross-sectional study. Nevertheless, taken together these findings may have significance for monitoring disease activity in MTB-infected individuals.

Conclusions
Our detailed dissection of the effects of HIV-TB co-infection on the MTB-specific CD4+ T-cell immune response has provided insight into clinically important features and indicated potentially useful immune correlates for monitoring disease activity. Our data accords with mounting evidence that MTB-specific T-cells secreting IL-2 with or without IFN-γ have a protective role to play in immune containment and suggests that the dual-secreting cells may be of greatest significance in this capacity. Secondly we identified a correlate of exhausted and dysregulated antimycobacterial immunity in active TB with advanced HIV co-infection that may partially explain certain characteristic clinical features of this life-threatening condition. Our hypothesis is that the immunophenotype (cytokine secretion profile and expression of PD-1) of the MTB-specific CD4+ T-cell response reflects the manifest failure of MTB containment in HIV co-infection i.e. active TB. Clinical observations suggest that in the presence of HIV infection this failure is profound with widespread dissemination of MTB. This is indicated by both the loss of CD4+ IFN-γ and IL-2 secreting cells and the increased frequency of PD-1 expression, which we observed in TB and HIV co-infection. If large longitudinal studies can prospectively validate our findings, these correlates may help to pave the way for the development and evaluation of new interventions to augment immune containment in LTBI and reverse immune dysregulation in active TB.
Supporting Information S1 Fig. Boolean gating of MTB-specific CD4+ functional cell subsets. PBMCs from donors were left unstimulated or stimulated with PPD or MTB-specific peptides for 16 hours with Monensin added after 2 hours. The cells were stained with a dead cell discriminator then fluorescence-conjugated antibodies against phenotypic markers before fixing, permeabilising and staining with antibodies targeting intracellular IFN-γ, IL-2 and TNF-α. Plots are representative of one individual with active TB/HIV and show PBMCs stimulated with PPD. Single, live, CD3 + lymphocytes were gated for CD4+ and CD8+ cells (top row). Gating was performed in two dimensions for quadrants containing IFN-γ, IL-2 and TNF-α positive and negative cells, (middle row). Three-dimensional Boolean gating was used to select cytokine subsets; an example of CD3+CD4+ IFN-γ-only-secreting cells is shown (bottom row) selecting only IFN-γ-onlysecreting cells from quadrants in the middle row (Q1, Q5 and Q12). Multi-dimensional Boolean gating allows the user to exclusively select subsets secreting specific cytokines, which is represented 2-dimensionally here. For example selecting IFN-γ-only-secreting cells excludes all cells secreting IL-2 or TNF-α. Live, CD3+ lymphocytes were gated for CD4+ and CD8+ cells, then for quadrants containing IFN-γ, IL-2 and TNF-α positive and negative cells, Boolean gating was used to select cytokine subsets in 3 dimensions to generate 7 cytokine subsets, an example of CD3+CD4+ IFN-γ-only secreting cells is shown and PD-1 expression was determined using PerCP-Cy5.5 fluorescence minus one as a gating control. (EPS)