The frequency of interleukin‐1β‐producing monocytes is significantly associated with varicella‐zoster responses of nursing home residents

Summary

Previous studies have demonstrated that the status of the T cell compartment and inflammation‐related factors are associated with the immunogenicity of the varicella‐zoster virus (VZV) vaccine in older adults; however, little is known about the roles of other immune cell subsets known to influence the generation and maintenance of immunological memory. Responses to a live‐attenuated VZV vaccine were studied in relation to peripheral blood mononuclear cell (PBMC) composition and function in a sample of 30 nursing home residents (aged 80–99 years). Interferon‐gamma enzyme‐linked immunospot (ELISPOT) was used to measure VZV responses at baseline and 6 weeks following vaccination, and associations were sought with the frequencies of monocytes and T, B and natural killer (NK) cells and the production and secretion of cytokines following their ex‐vivo stimulation with different agents. While only the frequency of interleukin (IL)‐6⁺ CD14⁺ monocytes was inversely associated with post‐vaccination VZV response, amounts of IL‐1β, IL‐10, IL‐17A and tumour necrosis factor (TNF) secreted by PBMCs and the frequency of IL‐1β⁺ CD14⁺ monocytes was positively correlated with pre‐vaccination VZV response. Furthermore, both bivariate correlation and causal mediation analyses supported the notion that IL‐1β⁺ CD14⁺ monocytes were significant mediators of the associations between IL‐1β and TNF secretion by PBMCs and pre‐vaccination VZV responses. Our findings implicate a strong cytokine response mediated by inflammatory IL‐1β⁺ monocytes in coordinating responses of long‐lived VZV‐reactive memory T cells, but with an opposing effect of IL‐6⁺ CD14⁺ monocytes. Whether monocyte status promotes or inhibits the induction and/or maintenance of these memory T cells later in life has yet to be determined.

In summary, the frequency of IL‐6⁺ monocytes and IL‐10 secreted by PBMCs was identified as potential predictive biomarkers of VZV vaccine immunogenicity (VZV‐responses following vaccination). We have also provided evidence that the pro‐inflammatory environment supported by IL‐1β⁺ monocytes was associated with the VZV‐response prior to vaccination in nursing home residents, and that these monocytes might also play a role in the induction and/or maintenance of VZV‐specific memory T cells acquired following the first exposure to VZV (probably during childhood).

Introduction

Primary infection with varicella‐zoster virus (VZV) generally occurs during childhood, leading to chickenpox. Following resolution of primary infection, VZV establishes latency in dorsal root ganglia, where it persists indefinitely. Reactivation and subsequent replication of the virus results in herpes zoster (HZ), a painful, neurocutaneous disorder commonly known as shingles [1]. Although the incidence of HZ has been estimated to be 4·47 ± 0·3 per 1000 person‐years across all ages, this rises rapidly after 50 years of age, reaching up to 10 cases per 1000 person‐years among those 80 years of age and older [2, 3]. Adults older than 60 years are at an increased risk of experiencing post‐herpetic neuralgia (PHN), the most common disabling complication of HZ, defined as chronic pain persisting beyond 90 days following the onset of rash [1].

VZV‐specific memory T cells acquired following the first exposure to the virus play an essential role in preventing VZV reactivation. However, this specific immunity declines with age, and both the frequency and intensity of VZV‐specific T cell responses have been shown to be inversely correlated with the incidence and severity of HZ, as well as with post‐herpetic neuralgia (PHN) [4, 5]. The first vaccine designed to prevent HZ in older adults, a live attenuated vaccine based on the Oka/Merck strain of VZV (Zostavax; Merck, Darmstadt, Germany), has been shown to be effective in stimulating the proliferation of VZV‐specific memory CD4 and CD8 T cells in adults older than 55 years [6], while reducing the incidence of HZ and PHN by more than 50% [7]. However, while this vaccine affords protection to approximately 70% of adults aged 50–59 from HZ, its efficacy progressively decreases with age, falling to 18% in those aged 80 years and older [7, 8, 9]. A newer adjuvanted non‐live recombinant vaccine (Shingrix; GlaxoSmithKline, Brentford, UK) has been shown to have a greater efficacy in preventing shingles, particularly in adults aged 80 years and older [10, 11]; however, responses have also been shown to decline with age [12].

These observations are in accordance with other studies that have reported a negative impact of age on the persistence of VZV‐specific T cell responses after vaccination [13, 14]. We have previously shown that the number of VZV‐specific interferon (IFN)‐γ‐secreting cells both before and after VZV vaccination were significantly reduced in nursing home residents aged 80–102 years relative to community‐dwelling older adults aged 60–75 years. This was despite the fact that the expansion of VZV‐specific T cells in response to the vaccine was similar in both cohorts, suggesting a potential role for other immune factors in the decline of vaccine effectiveness with age [15]. Humoral immunity may also be important and wane with age, although most studies have focused on T cell immunity [16]. Weinberg et al. showed that CD8 T cells responding to VZV in older adults predominantly expressed CD57 and programmed cell death 1 (PD‐1) (CD279), markers that have been associated with immune senescence and exhaustion, respectively, and that the frequencies of both CD4 and CD8 T cells with this phenotype were inversely correlated with VZV‐specific memory and effector T cell responses [17]. We have further shown that higher levels of cytomegalovirus (CMV)‐specific CD4 T cells, regulatory T cells (T_regs) and serum C‐reactive protein (CRP), a marker of chronic inflammation, prior to vaccination, correlate with decreased immunogenicity of VZV vaccine in nursing home residents [15, 18].

While these studies have greatly improved our understanding of how T cells and inflammation‐related factors affect the immunogenicity of the VZV vaccine in older adults, other potential mediators have yet to be investigated. For example, other immune subsets such as natural killer (NK) cells and B cells have been found in ganglia of donors following VZV reactivation [19], while cytotoxic NK cells with adaptive immune memory‐like features were recruited to sites of VZV skin test antigen challenge in VZV‐experienced volunteers, suggesting an involvement of these cells in VZV immunity [20]. Furthermore, although little is known concerning the role of monocytes in VZV immunity or vaccine responses in older adults, recent work has shown that human monocytes are permissive to VZV infection, and therefore may play a role controlling within‐host dissemination and the maintenance of effective immunological memory [21]. Given this, the primary objective of the present study was to investigate which other immune factors were associated with VZV vaccine immunogenicity in nursing home residents. Specifically, we investigated the frequencies of monocytes, T, B and NK cells, and the intracellular production and secretion of cytokines following ex‐vivo stimulation with Toll‐like receptor (TLR) ligands [lipopolysaccharide (LPS) and Pam3CysSerLys4 (PAM₃CSK₄)] and other immune agonists [phosphate‐buffered saline (PBS) as vehicle, soluble anti‐human CD3 and CD28 monoclonal antibodies (sCD3/sCD28) and phytohaemagglutinin (PHA)].

Materials and methods

Participants and study design

Participants in the current study were drawn from a previously published safety and efficacy trial of the live‐attenuated VZV vaccine (Zostavax; clinical trial registration: NCT01328548) in nursing home residents aged 80–102 [15], and have been studied in follow‐up reports [18, 22]. From this cohort, a convenience sample of 32 participants whose distribution of VZV vaccine responses and demographics [i.e. age, sex, frailty, cytomegalovirus (CMV) reactivity] were similar to that of the entire cohort were selected for the current analysis (Table 1). Vaccine immunogenicity data were not complete for one participant, and peripheral blood mononuclear cells (PBMCs) from a second participant failed to respond to ex‐vivo stimulation; hence, the final sample of participants included in the current study numbered 30. Written informed consent was obtained from all participants and the study protocol and consent procedures were performed according to the Declaration of Helsinki and were approved by the McMaster Research Ethics Board.

Table 1.

Summary of participants included in this study and that of the entire original cohort

		Current subset	Entire cohort
		(n = 30)	(N = 188)
Demographics
Age		88 (86–91)	89 (85·5–92)
Missing		0 (0%)	1 (0·5%)
Sex
Female		26 (86·7%)	153 (81·4%)
Male		4 (13·3%)	35 (18·6%)
Cytomegalovirus status
No		5 (16·7%)	44 (23·4%)
Yes		25 (83·3%)	142 (75·5%)
Missing		0 (0%)	2 (1·1%)
Frailty index		0·31 (0·271–0·379)	0·31 (0·257–0·371)
Missing		3 (10·0%)	41 (21·8%)
VZV responses
Baseline (pre‐Vax)	Raw	28·3 (18–45·1)	28·7 (24·3–34·1)
	Ln	3·3 (2·86–3·81)	3·4 (3·19–3·53)
Follow‐up (post‐Vax)	Raw	57·8 (40·9–81·1)	55·6 (47·9–64·6)
	Ln	4·1 (3·71–4·39)	4 (3·88–4·16)
Fold‐change	Raw	2·04 (1·39–3·01)	1·93 (1·69–2·23)
	Ln	0·71 (0·344–1·1)	0·66 (0·524–0·801)

Age and frailty are median (interquartile range) and the VZV responses are the geometric mean and 95% confidence interval, either the raw value or natural‐log (Ln).

Blood collection and PBMC preparation

Briefly, blood samples were collected prior to and 6 weeks following subcutaneous vaccination, and PBMCs were cryopreserved in 10% dimethylsulphoxide (DMSO)/human AB serum in the vapour phase of liquid nitrogen. Subsequently, PBMCs from the baseline collection were thawed at 37°C in prewarmed X‐VIVO 10 medium (Lonza, Basel, Switzerland) supplemented with 5% human AB serum (Corning, New York, NY, USA) and resuspended at 2 × 10⁶/ml in the same medium in four batches on different days over a period of 2 weeks. For each batch, PBMCs were separated into two plates and stimulated to measure either cytokine secretion or intracellular cytokine production; a third plate of unstimulated cells was prepared to measure cellular immunophenotype (i.e. cell surface staining, below).

IFN‐g enzyme‐linked immunospot (ELISPOT) assay

IFN‐γ ELISPOT was performed to assess VZV vaccine responses, as previously described [18]. Briefly, cryopreserved PBMCs from baseline and 6 weeks post‐vaccination collections were thawed at 37°C in prewarmed X‐VIVO 10 medium (Lonza) supplemented with 5% human AB serum (Corning) and resuspended in the same medium. A total of 5 × 10⁵ PBMCs was added to each well of multi‐screen‐IP membrane plates (Millipore, Burlington, MA, USA) coated with anti‐human recombinant IFN‐γ monoclonal antibody. PBMCs were stimulated with lysed VZV‐infected MRC‐5 cells or CMV‐Epstein–Barr virus (EBV)‐influenza (CEF) peptide pools (5 μg/ml) for 18 h at 37°C/5% CO₂. Plates were subsequently washed and incubated with a biotinylated anti‐human recombinant IFN‐γ antibody, followed by incubation with streptavidin–alkaline phosphatase allowing spot development. Spot‐forming cells were enumerated using the CTL ImmunoSpot Image Analyzer system and counting software. VZV responses prior to and after vaccination were abbreviated as pre‐ and post‐Vax, respectively, and vaccine immunogenicity was determined by calculating the ratio of VZV response after vaccination to VZV response prior to vaccination (i.e. baseline). VZV and CEF responses at each time‐point are presented as the natural‐log.

Cell surface staining for immunophenotyping

PBMCs from the baseline collection were centrifuged and resuspended in a 1 : 1 solution of human AB serum and fluorescence activated cell sorter (FACS) wash [PBS/2·5% bovine serum albumin (BSA)/10 mM ethylenediamine tetraacetic acid (EDTA)], the latter containing the following fluorophore‐conjugated antibodies: CD14‐phycoerythrin (PE)‐Dazzle, CD16‐Alexa647, human leucocyte antigen D‐related (HLA‐DR)‐fluorescein isothiocyanate (FITC), CD3‐allophycocyanin‐cyanin 7 (APC‐Cy7), CD4‐peridinin chlorophyll (PerCp)‐Cy5.5, CD8‐PE‐Cy7, CD45‐Brilliant Violet510, CD56‐PE, NKp46‐PE (CD335‐PE) and CD19‐Alexa700 (BD Biosciences, Biolegend, eBioscience, San Diego, CA, USA). Following a 300‐min incubation at room temperature, PBMCs were washed once in FACS wash and fixed for 10 min at room temperature with 2% paraformaldehyde; cellular fluorescence was measured using a BD LSR II flow cytometer. The frequencies of HLA‐DR⁺ monocytes (classical, CD14⁺⁺CD16⁻; intermediate, CD14⁺⁺CD16⁺; non‐classical, CD14⁺CD16⁺), CD3⁺CD4 and CD8 T cells, NK cells (CD56⁺ and/or NKp46⁺) and CD19⁺ B cells, relative to CD45⁺ PBMCs, were estimated using the analysis software FlowJo (TreeStar, Inc., Ashland, OR, USA). The gating strategy is presented in Supporting information, Fig. S1.

Cytokine secretion analysis

PBMCs from the baseline collection were treated with PBS (mock), 25 ng/ml LPS, 250 ng/ml Pam₃CSK₄, 5 µg/ml each of sCD3/sCD28 or 2·5 µg/ml PHA for 24 h at 37°C/5% CO₂. These treatments were chosen in order to model constitutive cytokine secretion (i.e. PBS) and induce primarily myeloid (i.e. TLR agonists, LPS and Pam₃CSK₄) and T cell responses (i.e. sCD3/sCD28 and PHA), although both direct and indirect induction of other cell types from these agonists is also possible. Following treatment, supernatants were frozen at −20°C before being thawed to measure the concentrations of IL‐1β, IL‐10, IL‐12, IL‐17A and tumour necrosis factor (TNF) by multiplexed bead enzyme‐linked immunosorbent assay (ELISA) (Millipore), according to the manufacturer’s instructions.

Intracellular cytokine staining

PBMCs from the baseline collection were treated with a protein transport inhibitor cocktail (i.e. Golgi plug; eBioscience) and challenged with PBS (mock), a combination of 25 ng/ml LPS and 250 ng/ml Pam₃CSK₄ or 5 µg/ml each of sCD3/sCD28 for 4 h 37°C/5% CO₂. Following stimulation, PBMCs were resuspended in cold 2 mM EDTA for 10 min at 4°C, centrifuged and stored at 4°C in RPMI supplemented with 10% FBS. The following day, PBMCs were centrifuged and resuspended in a 1 : 1 solution of human AB serum and FACS wash, the latter containing the following conjugated antibodies: CD14‐PE‐Dazzle, CD3‐APC‐Cy7, CD4‐PerCp‐Cy5.5, CD8‐PE‐Cy7 and CD45‐Brilliant Violet510 (BD Biosciences, Biolegend, eBioscience). Following a 30‐min incubation at room temperature, PBMCs were resuspended in permeabilization buffer (eBioscience) and incubated for an additional 30 min at room temperature. PBMCs were then centrifuged and resuspended in permeabilization buffer containing IFN‐γ‐Pacific Blue, IL‐6‐FITC, TNF‐Alexa700 and IL‐1β‐PE (BD Biosciences, Biolegend, eBioscience) and incubated at room temperature for 30 min. Afterwards, PBMCs were washed, fixed with 2% paraformaldehyde and stored at 4°C until analysis. The percentage of cytokine‐positive CD14⁺ monocytes and CD3⁺ CD4 or CD8 T cells, relative to CD45⁺ PBMCs, were measured as described above. The cytokines chosen for analysis were those known to be secreted by monocytes and T cells within the time‐frame, as described above.

Statistical analysis

Cellular frequency and the frequency of cytokine positive‐cell subsets were summarized as the mean and 95% confidence interval (CI), while cytokine concentrations were summarized as the geometric mean and 95% CI; these measures were obtained from experiments performed only on baseline PBMCs. Associations of these baseline measures with natural‐log‐transformed IFN‐γ ELISPOT responses following VZV or CEF stimulation were performed using multiple linear regression. All models were adjusted for age and sex, and for those in which log‐transformed VZV responses of participants post‐vaccination were the outcome, log‐transformed VZV responses at baseline (i.e. pre‐vaccination) were also included as a covariate; this is the preferred approach for outcomes such as vaccine responses that tend to exhibit ceiling effects [23]. For models in which cellular frequency or cytokine responses (i.e. PBMC secretion or frequency of cytokine positive cells) to mock treatment were included as predictors, only a single predictor was included per model and no other covariates were included, other than those listed above. Models including cellular responses to an agonist (i.e. LPS, Pam3CSK4, sCD3/sCD28 or PHA) were similar to this, and only the response to mock treatment for that given cytokine was also included in order to account for the confounding effect of constitutive expression. As an example, models in which baseline VZV responses were regressed against age, sex and PBMC secretion of IL‐1β following LPS treatment also included the fixed effect of PBMC secretion of IL‐1β following mock treatment. Cytokine concentrations were natural‐log‐transformed prior to regression analysis in order to minimize the impact of extreme values; 0·01 was added to all values to facilitate transformation.

Bivariate correlations between natural‐log‐transformed secreted cytokine concentration and the percentage of IL‐1β‐producing monocytes was performed using Pearson’s correlation. For casual mediation analysis, we used the R ‘mediation’ package [24], which estimates the proportion of an effect by a treatment or exposure on a given outcome that occurs through a mediator, commonly known as the indirect effect. Specifically, the direct (i.e. average direct effect) and indirect (i.e. average causal mediation effect) effects are estimated by comparing the treatment effect obtained from a regression model that includes only the treatment variable to one that includes both the treatment and mediator variables (all relevant covariates are included in both models). For the current study, we considered PBMC cytokine secretion under various conditions as the treatment, VZV response prior to vaccination as the outcome and the frequency of IL‐1β⁺ CD14⁺ monocytes following mock treatment as the mediator. The proportion mediated and 95% CI is presented as a percentage, where 100% mediation indicates that the entire effect occurs through the mediator. There are no distributional assumptions for the CI, so an estimate of proportion mediated is significant if zero does not fall within [25]. All analyses were performed in the R environment (version 3.6).

Results

Frequencies of peripheral blood immune subsets and responses to ex‐vivo stimulation

Frequencies of peripheral immune cell subsets prior to vaccination did not deviate from expectations, being T cells (mean = 63% of CD45⁺ cells), monocytes (15%), NK cells (11%) and B cells (5%). Within the T cell subset, mean values were 42% CD4⁺ and 20% CD8⁺ of CD45⁺ cells. Monocytes were mainly of the ‘classical’ phenotype (12·4% of CD45⁺ cells), followed by intermediate (1·5%) and non‐classical monocytes (0.8%) (Fig. 1a).

Fig. 1.

Cellular frequencies of immune subsets and cytokine secretion following stimulation of peripheral blood mononuclear cells (PBMCs) in nursing home residents. (a) Frequencies of T cells, monocytes, B cells and natural killer (NK) cells among mononuclear cells (left panel); frequencies of CD4 and CD8 T cells among overall T cells (middle panel) and frequencies of classical, intermediate and non‐classical monocytes among the whole monocyte population (right panel) measured in the peripheral blood of 30 nursing home residents. Means with 95% confidence intervals (CIs) are represented on the graphs. (b) Levels of interleukin (IL)‐10, IL‐12, IL‐17A, IL‐1β and tumour necrosis factor (TNF) secreted by PBMCs following stimulation with mock [phosphate‐buffered saline (PBS)], sCD3/sCD28, phytohaemagglutinin (PHA), lipopolysaccharide (LPS) and Pam3CysSerLys4 (PAM₃CSK₄) in 30 nursing home residents. Geometric means with 95% CIs are represented on the graphs.

Following ex‐vivo stimulation of these baseline pre‐vaccination PBMCs with different agonists, TNF was found to be the most highly secreted cytokine, regardless of the treatment, followed by IL‐10 and IL‐1β, whereas amounts of IL‐17A and IL‐12 remained very low. The T cell mitogen PHA was the most potent agonist, inducing the highest concentrations of all cytokines secreted by PBMCs. Soluble CD3/CD28 (T cell stimulus) and LPS (TLR‐4 agonist) induced similar levels of cytokines, while supernatants contained only low levels of cytokines following PAM₃CSK₄ (TLR‐1/2 agonist), similar to the PBS control (Fig. 1b).

Cytokines secreted by PBMCs are broadly associated with VZV responses at baseline but not after vaccination

Using linear regression, no significant correlations were observed between cellular frequencies and VZV responses before or after vaccination (Supporting information, Table S1). In addition, of all the cytokines produced by PBMCs, only IL‐10 secreted following LPS stimulation was found to be associated (negatively) with post‐vaccination VZV responses (β = 0·294, P < 0·05) (Fig. 2a, Supporting information, Table S2). In contrast, significant positive associations were observed between secreted cytokines and VZV responses prior to vaccination, where the natural‐log pre‐Vax was observed to increase for every 1 standard deviation change in TNF (β = 0·609, P < 0·05), IL‐10 (β = 0.492, P < 0·05) and IL‐1β (β = 0·566, P < 0·05) following mock treatment and IL‐17A following LPS (β = 0·728, P < 0·05) and sCD3/sCD28 treatments (β = 0·708, P < 0·01) (Fig. 2a, Supporting information, Table S2). Unlike VZV responses, cytokine secretion by PBMCs was not correlated with CEF responses (i.e. positive control) before or after vaccination (Supporting information, Table S3). Thus, these data suggest that VZV responses prior to, but not following, vaccination are broadly associated with PBMC cytokine secretion at baseline, and VZV responses at either time‐point are not associated with PBMC cell composition.

Fig. 2.

Associations of peripheral blood mononuclear cells (PBMCs) cytokine secretion to varicella zoster virus (VZV) responses before (pre‐Vax) and after (post‐Vax) vaccination. Association between natural‐log‐transformed (a) VZV responses post‐Vax or (b) VZV responses pre‐Vax with standardized (mean = 0, standard deviation = 1) levels of interleukin (IL)‐10, IL‐12, IL‐17A, IL‐1‐β and tumour necrosis factor (TNF) secreted by PBMCs following stimulation with mock [phosphate‐buffered saline (PBS)], lipopolysaccharide (LPS), Pam3CysSerLys4 (PAM₃CSK₄), sCD3/sCD28 and phytohaemagglutinin (PHA). The β coefficients and 95% confidence intervals are represented on the graphs. Points above the dotted red line indicate a positive correlation with VZV responses post‐Vax or VZV responses pre‐Vax, and points below indicate and inverse correlation; results are considered significant when a set of error bars do not cross this line (*P < 0·05, **P < 0·01).

The frequency of IL‐1β+ CD14+ monocytes positively correlates with VZV responses prior to vaccination

In order to elucidate the potential cellular mediator of cytokines that correlated with pre‐vaccination VZV responses, we performed intracellular cytokine analysis. PBMCs were treated with mock, sCD3/sCD28 and LPS⁺PAM₃CSK₄ and frequencies of CD14⁺ monocytes producing IL‐1β, IL‐6 or TNF, and CD4⁺ or CD8⁺ T cells producing IFN‐γ or TNF were assessed. As expected, high frequencies of IL‐1β⁺, IL‐6⁺ and TNF⁺ CD14⁺ monocytes were observed following LPS⁺PAM₃CSK₄ stimulation, while relatively low frequencies of IL‐6⁺ and TNF⁺ monocytes were seen following mock and sCD3/sCD28 treatments (< 5%) (Fig. 3a). The frequency of IL‐1β⁺ monocytes following sCD3/sCD28 stimulation, although relatively high (47%), was only slightly higher than when mock treatment was applied (36·3%) (Fig. 3a). Stimulation with sCD3/sCD28 induced the highest production of IFN‐γ and TNF by T cells, but the frequency of IFN‐γ⁺ and TNF⁺ CD4 T cells remained low relative to the frequency of IFN‐γ⁺ and TNF⁺ CD8 T cells (Fig. 3a).

Fig. 3.

Cytokine production by CD14⁺ monocytes and CD4 and CD8 T cells and their relationships to varicella zoster virus (VZV) responses prior to and after vaccination. (a) Frequencies of interleukin (IL)‐1β⁺, IL‐6⁺ and tumour necrosis factor (TNF)⁺ CD14⁺ monocytes following mock [phosphate‐buffered saline (PBS)], sCD3/sCD28 and lipopolysaccharide (LPS)+Pam3CysSerLys4 (PAM₃CSK₄) treatments (left panel); frequencies of interferon (IFN)‐γ⁺ and TNF⁺ CD4 (middle panel) and CD8 (right panel) T cells following mock and sCD3/sCD28 treatments in 30 nursing home elderly residents. Means with 95% confidence intervals (CIs) are represented on the graphs. (b) The Ln VZV response and (c) Ln cytomegalovirus (CMV)‐Epstein–Barr virus (EBV)‐influenza (CEF) response prior to (pre‐Vax) and after vaccination (post‐Vax) correlated with frequencies of IL‐1β⁺ CD14⁺ monocytes and TNF⁺ CD4 T cells following mock treatment; IFN‐γ⁺ CD8 T cells stimulated with sCD3/sCD28 and IL‐6⁺ CD14⁺ monocytes stimulated with LPS⁺PAM₃CSK₄ in 30 nursing home elderly residents (from left to right). Slope (β), 95% CIs and significance were derived from adjusted linear regression models (*P < 0·05, **P < 0·01).

Of the frequencies of cytokine‐producing cells mentioned above (Fig. 3a), only the frequency of IL‐6⁺ CD14⁺ monocytes was found to be (negatively) associated with VZV response after vaccination. Multiple regression analysis revealed that natural‐log VZV responses post‐Vax decreased 0·04 with every 1% increase in IL‐6⁺ CD14⁺ monocytes following LPS⁺PAM₃CSK₄ treatment (P < 0·05) (Fig. 3b, Supporting information, Table S4). In contrast, there was a significant positive association between VZV responses prior to vaccination and the frequency of IL‐1β⁺ CD14⁺ monocytes. Specifically, natural‐log VZV responses pre‐Vax increased 0·04 with every 1% increase in IL‐1β⁺ CD14⁺ monocytes following mock treatment (P < 0·01) (Fig. 3b, Supporting information, Table S4); no associations with VZV responses pre‐Vax were observed for IL‐6⁺ or TNF⁺ CD14⁺ monocytes or for IL‐1β⁺ CD14⁺ monocytes under any other treatment condition (Supporting information, Table S4). The anticipated significant positive correlation between IFN‐γ⁺ CD8 T cells following sCD3/sCD28 stimulation and VZV responses pre‐Vax was indeed observed (β = 0·105, P < 0·05), while TNF⁺ CD4 T cells also exhibited a significant, but inverse, correlation with VZV responses pre‐Vax (β = −12·2, P < 0·01) (Fig. 3b, Supporting information, Table S4). Again, there were no associations between any of these populations and response to CEF except the TNF⁺ CD4 T cell frequency at baseline, which was inversely correlated with CEF prior to (β = −10·6, P < 0·05) and after (β = −10·1, P < 0·05) vaccination (Fig. 3c, Supporting information, Table S5). Thus, these data suggest that the frequency of IL‐6⁺ monocytes under LPS⁺PAM₃CSK₄ treatment at baseline is predictive of a decreased immunogenicity of the VZV vaccine (i.e. VZV responses after vaccination) while the frequency of IL‐1β⁺ monocytes in the absence of any acute stimulation at baseline significantly correlates with VZV responses prior to vaccination.

The frequency of IL‐1β+ CD14+ monocytes correlates with the IL‐1β and TNF response of PBMCs following ex‐vivo stimulation at baseline

Bivariate analysis was performed in order to more clearly understand the inter‐relationships among immune parameters that were found to be associated with the VZV response prior to vaccination; namely, the correlation between IL‐1β⁺ CD14⁺ monocyte frequency and PBMC cytokine secretion. The frequency of IL‐1β⁺ CD14⁺ monocytes following mock treatment was significantly correlated with actual IL‐1β secretion by PBMCs following stimulation with all agonists (P < 0·01) except LPS, as were TNF concentrations following mock (P < 0·01) and PHA treatments (P < 0·05) (Fig. 4). The secretion of IL‐10 following mock treatment was also significantly correlated with the level of IL‐1β⁺ CD14⁺ monocytes following mock treatment (P < 0·05) (Fig. 4). As these findings suggest that IL‐1β⁺ CD14⁺ monocytes mediate the secretion of IL‐1β, TNF and IL‐10 under various different treatment conditions, we performed causal mediation analysis. Indeed, in models with VZV responses prior to vaccination as the outcome, the frequency of IL‐1β⁺ CD14⁺ monocytes following mock treatment at baseline was found to mediate a significant proportion of the effect of PBMC IL‐1β secretion following mock treatment [proportion mediated (69%, 95% CI = 3.3, 272)], PAM₃CSK₄ (69%, 95% CI = 3.6, 267) and sCD3/sCD28 stimulation (67%,95% CI = 1.8, 256) and of TNF secretion following mock treatment (46%, 95% CI = 4.3, 152) at baseline. The proportion of the effect mediated by IL‐1β⁺ CD14⁺ monocytes for associations with IL‐1β and TNF following PHA and IL‐10 following mock treatment were all greater than 45%, but were not statistically significant. As a whole, these findings suggest that those PBMC cytokine responses that are strongly correlated with VZV responses prior to vaccination are probably mediated by IL‐1β⁺ monocytes.

Fig. 4.

Heat‐map showing the Pearson correlation coefficients between the frequency of interleukin (IL)‐1β⁺ CD14⁺ monocytes following mock [phosphate‐buffered saline (PBS)] treatment and peripheral blood mononuclear cell (PBMC) cytokine secretion [tumour necrosis factor (TNF), IL‐1β, IL‐17A, IL‐12 and IL‐10] according to stimulation conditions [mock, lipopolysaccharide (LPS), Pam3CysSerLys4 (PAM₃CSK₄), sCD3/sCD28 and phytohaemagglutinin (PHA)] (*P < 0·05, **P < 0·01, *** P < 0·001).

Discussion

The primary objective of our study was to identify novel cellular and cytokine markers of VZV responses in nursing home residents at baseline, in order to more clearly understand the correlates of VZV immunity prior to and following vaccination [15, 18]. Only two parameters were found to be associated with post‐vaccination VZV response; namely, IL‐10 secreted by PBMCs following LPS stimulation (positive correlation) and the frequency of IL‐6⁺ CD14⁺ monocytes following LPS stimulation (inverse relationship). More factors were found to be associated with the pre‐vaccination VZV response, in particular the secretion of IL‐10, IL‐1β, TNF and IL‐17A by PBMCs, and the frequency of IL‐1β⁺ CD14⁺ monocytes following mock treatment that all exhibited positive correlations. Finally, we found that the composition of the peripheral blood immune compartment, i.e. the frequencies of monocytes and T, B and NK cells, was not associated with VZV responses either before or after vaccination. Of note, there were no associations between immune parameters and the multi‐antigen‐positive control CEF response prior to and after vaccination, suggesting that the relationships we observed were probably VZV‐specific and not due to broad functional differences in IFN‐γ responses.

The treatments employed in our study were chosen due to their known ability to primarily stimulate monocytes (i.e. LPS, PAM₃CSK₄) or T cells (i.e. sCD3/sCD28, PHA). As expected, sCD3/sCD28, which has been shown to stimulate T cells comparably to immobilized antibodies [26], induced the production of IFN‐γ and TNF by both CD4 and CD8 T cells. Interestingly, this treatment seemed to also stimulate monocytes, as the frequency of IL‐1β⁺ monocytes following stimulation with sCD3/sCD28 was slightly higher than mock‐treated cultures. Although previous work has shown that soluble monoclonal antibodies can induce monocyte activation directly through the cross‐linking of cell‐surface Fcγ receptors [27, 28], it is most probable that the stimulation seen here occurred primarily through indirect mechanisms (e.g. via cytokines secreted by activated T cells). The activation of monocytes by the combined LPS and PAM₃CSK₄ treatment was significantly higher than that caused by sCD3/sCD28, which is as expected for TLR agonists. For associations with the post‐vaccination VZV response, only the frequency of IL‐6⁺ CD14⁺ monocytes following LPS⁺PAM₃CSK₄ stimulation at baseline was found to be significant, but this was an inverse relationship. IL‐6 is a proinflammatory cytokine known to be involved in numerous chronic inflammatory diseases [29] and its circulating levels significantly increases with age [30], especially so in nursing home residents, as we have shown [18]. Consistent with the data presented herein, plasma levels of IL‐6 were found to be inversely associated with hepatitis B virus (HBV) antibody responses to HBV vaccine in hepatitis C virus‐infected participants, suggesting an adverse effect of IL‐6 on vaccine responses [31]. In addition, we previously found that the level of serum CRP, another marker for inflammation, was inversely related to post‐vaccination VZV response in the same cohort of nursing home residents [18]. Interestingly, IL‐6 is required for the induction of CRP [32], supporting not only a strong relationship between these two inflammatory factors but also the idea that inflammation may be detrimental to vaccine responses.

IL‐10 secreted by PBMCs following LPS treatment at baseline was the one other parameter associated with post‐vaccination VZV responses. Our data are not in line with other studies that showed that high constitutive levels of IL‐10‐impaired live‐attenuated VZV vaccine immunogenicity, identified by an inverse relationship between IL‐10 and VZV response in elderly adults aged 60–80 years [33]. In addition, similar findings were reported from a cohort of older adults aged 63–85 years who were vaccinated against influenza, where high PHA‐induced IL‐10 production by leucocytes was associated with a low response to the vaccine [34]. Further investigations will be necessary to understand our discrepant results.

Although we were primarily interested in estimating associations with responses to VZV vaccination, we found instead that the secretion of cytokines by PBMCs at baseline, namely IL‐1β, IL‐10, IL‐17A and TNF, were positively associated with VZV responses prior to vaccination, as was the frequency of IL‐1β⁺ CD14⁺ monocytes following mock treatment at baseline. Additional analyses indicated that the frequency of IL‐1β⁺ CD14⁺ monocytes was strongly correlated with the secretion of IL‐1β, and to a lesser extent TNF and IL‐10, following PBMC stimulation, and that IL‐1β⁺ CD14⁺ monocytes were a significant mediator of the secretion of these cytokines. Taken together, these data suggest an involvement of inflammatory monocytes in the induction or maintenance of VZV‐specific memory T cells of older adults through the secretion of IL‐1β.

Studies have shown that IL‐1β‐producing myeloid cells (i.e. monocytes, macrophages and dendritic cells) are critical for the activation of antigen‐specific T cells [35] required to mount an appropriate response to a pathogen. Following resolution of infection, antigen‐specific effector T cells undergo clonal contraction leading to the death of short‐lived effector T cells and survival of long‐lived memory T cells [36] which localize throughout the body, mainly in the bone marrow [37, 38]. Recently, it was shown that inflammatory monocytes may play an important role in the maintenance of these cells. Thus, in a mouse model of the CD8 T cell response to respiratory infection with vaccinia virus, it was found that deficiency of CCR2 severely impaired the recruitment of Ly6C⁺ CD11b⁺ inflammatory monocytes to sites of inflammation/infection [39]. Interestingly, while similar frequencies of antigen‐specific CD8 T cells were observed in both C‐C chemokine receptor type 2 (CCR2^−/−) and wild‐type (WT) mice at day 8 post‐vaccination, lower frequencies of antigen‐specific memory CD8 T cells were found in CCR2^−/− mice at day 50 post‐infection [39]. This suggests that inflammatory monocytes were required for the persistence of long‐lived memory CD8 T cells, but not for CD8 T cell expansion, during infection. Our data are consistent with this because frequencies of IL‐1β⁺ CD14⁺ monocytes, probably corresponding to the Ly6C^high inflammatory monocyte populations in mice, were positively correlated with VZV responses prior to vaccination (i.e. VZV‐specific long‐lived memory T cells), but not with VZV responses following vaccination (i.e. post‐vaccination generated VZV‐specific T cells).

Other studies have demonstrated the involvement of IL‐1β in the activation of antigen‐specific T cells and induction and/or maintenance of memory T cells, albeit without explicitly investigating the role of monocytes or other myeloid cells, the predominant producers of the cytokine [40]. Indeed, the proliferative recall responses of methyl BSA (mBSA)‐specific T cells derived from IL‐1β^−/− mice were significantly reduced compared to control mice, suggesting a critical role of IL‐1β in the development of delayed‐type hypersensitivity responses against mBSA [41]. Other work has shown that the injection of IL‐1β into mice immunized with ovalbumin (OVA) not only increased the frequency and effector functions of OVA‐specific CD8 T (OT‐I) cells, but also enhanced memory responses of the OT‐I cells [42]. A similar role of IL‐1β has been reported by the same investigators in terms of the enhancement of expansion and memory responses with regard to CD4 T cells [43]. Finally, the frequency of OT‐I cells expressing CD127 (IL‐7Rα) isolated from lymph nodes, spleen, liver and lungs of mice injected with OVA was found to be significantly higher in mice treated with IL‐1β than in those treated with vehicle [44]. Given that CD127 has been identified as a marker of long‐lived memory T cells and is consequently strongly associated with T cell survival [45], future studies into the role of inflammatory monocyte‐derived IL‐1β on CD127 induction would be of particular interest.

Our study has some limitations. First, immune parameters were measured in the peripheral blood of older adults but the skin is the major site involved in VZV reactivation; hence, while peripheral immune biomarkers may reflect what is happening in the skin, differences have also been noted in previous work [46]. Secondly, our sample size was relatively small, which limits generalizability. In addition, our study investigated responses to live‐attenuated VZV vaccine (Zostavax), which has been replaced by a non‐live recombinant VZV vaccine (Shingrix) as the currently recommended vaccine for older adults. Nonetheless, this does not discount the observations we observed pre‐vaccination, and we believe that our post‐vaccination findings are still relevant from a vaccine immunology perspective. Lastly, the phenotype and function of immune cell subsets was assessed on cryopreserved PBMCs which could have impacted resultant frequencies of cells, in particular for CD16‐expressing monocyte subsets [47]. That said, in a previous paper we assessed frequencies of monocyte subsets in fresh blood of a similar cohort of nursing home elderly residents aged 81–100 years. In the previous and present study, the mean frequency of CD14⁺⁺CD16⁺ intermediate monocytes was 0·58 versus 1·5%, respectively, while the frequency of CD14⁺CD16⁺ non‐classical monocytes was 0.55 versus 0.8%, respectively. Unfortunately, we are unable to evaluate the impact of cryopreservation on the observed cytokine production following treatment; however, a recent study shows that frequencies of IL‐1β⁺, IL‐6⁺ and TNF‐α⁺ monocytes following LPS stimulation were similar in fresh and frozen blood [48].

In summary, the frequencies of IL‐6⁺ monocytes and IL‐10 secreted by PBMCs were identified as potential predictive biomarkers of VZV vaccine immunogenicity (VZV responses following vaccination). We have also provided evidence that the proinflammatory environment supported by IL‐1β⁺ monocytes was associated with the VZV response prior to vaccination in nursing home residents, and that these monocytes might also play a role in the induction and/or maintenance of VZV‐specific memory T cells acquired following the first exposure to VZV (probably during childhood). Future studies will be needed to elucidate whether or not IL‐1β‐producing monocytes are causally related to the induction and maintenance of VZV‐specific memory, and if so, the underlying mechanisms responsible.

Disclosures

J. E. M.’s institution has received consulting fees for her participation in advisory boards or data safety monitoring boards for Sanofi, GSK, Merck, RestorBio and Medicago.

Author contributions

C. P. V. designed the research study and performed the experiments; C. P. V. and E. P. analyzed the data and drafted the original manuscript; C. P. V., E. P., D. M. E. B., J. E. M., G. P. and M. L. critically revised and edited the manuscript; C. P. V., E. P., D. M. E. B., J. E. M., G. P. and M. L. approved the submitted and final version of the manuscript.

Supporting information

Fig. S1. Gating strategy for the immunophenotype analysis. Duplicates were removed gating on FSC‐W vs. FSC‐A. Next, CD45⁺ leukocytes were selected and further separated into CD3⁺ and CD3^‐ cells. Expression of CD4 and CD8 was then used to identify CD4 and CD8 T cells, from CD3⁺ cells. The remaining CD3^‐ cells were separated into CD14^‐ and CD14⁺ cells. CD14⁺ HLA‐DR⁺ cells were selected and next, the expression of CD14 and CD16 was used to identify classical (CD14⁺⁺ CD16^‐), intermediate (CD14⁺⁺ CD16⁺) and non‐classical (CD14⁺ CD16⁺) monocytes. NK and B cells were identified from CD14^‐ cells as CD56/NKp46⁺ and CD19⁺ cells, respectively. CLS: classical; INT: intermediate; NON: non‐classical.

Table S1. Association of cellular frequency to VZV‐responses before (pre‐Vax) and after (post‐Vax) vaccination.

Table S2. Associations of PBMC cytokine secretion to VZV responses before (pre‐Vax) and after (post‐Vax) VZV vaccination.

Table S3. Associations of PBMC cytokine secretion to CEF responses before (pre‐Vax) and after (post‐Vax) VZV vaccination.

Table S4. Associations of cytokine production by CD14⁺ monocytes, CD4 and CD8 T cells to IFNγ‐ELISpot VZV‐responses before (pre‐Vax) and after (post‐Vax) vaccination.

Table S5. Associations of cytokine production by CD14⁺ monocytes, CD4 and CD8 T cells to IFNγ‐ELISpot CEF responses before (pre‐Vax) and after (post‐Vax) VZV vaccination.

Data Availability Statement

The data sets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.