cDC1 IL-27p28 production predicts vaccine-elicited CD8+ T cell memory and protective immunity

Abstract

While adjuvants and formulations are often either empirically derived, or at best judged by their ability to elicit broad inflammation, it would be ideal if specific innate correlates of adaptive immunity could be identified, in order to set a universally applicable benchmark for adjuvant evaluation. Using an IL-27 reporter transgenic mouse model, we show here that cDC1 IL-27 production in the draining lymph node 12 hours after subcutaneous vaccination directly correlates with downstream CD8+ T cell memory and protective immunity against infectious challenge. This correlation is robust, reproducible, predictive, entirely unique to vaccine biology, and is the only innate correlate of CD8+ T cell immune memory yet to be identified. Our results provide new insights into the basic biology of adjuvant-elicited cellular immunity and have clear implications for the screening and evaluation of novel adjuvants.

Introduction

Over the past several decades, our understanding of CD8 T cell biology and the formation of protective cellular-mediated immune memory has expanded greatly, driven significantly by robust murine models of infection such as Listeria monocytogenes (LM) and lymphocytic choriomeningitis virus (1, 2). These models have facilitated insights into how CD8 T cells encounter antigen, become activated, expand, and contract over time to form stable protective memory pools. It is reasonable to assume that the ‘rules’ of T cell behavior established in these systems may be applied toward understanding an immune response against an unknown infectious agent. Their applicability to subunit vaccination has become more questionable. Indeed, a growing body of research now suggests that key elements of our understanding of T cell biology derived from infectious models do not apply to subunit vaccine-elicited cellular immunity, and vice versa. For example, there is an obligate role for IL-27 (3, 4) and IL-15 (4) in facilitating maximal vaccine-elicited CD4 and CD8+ T cell responses to the majority of vaccine adjuvants, whereas the loss of either of these cytokines does not diminish the CD4 or CD8 T cell responses against infectious challenge (3, 5–8).

Not only does the biology of vaccine-elicited cellular immunity appear to diverge from that of infectious biology, it also seems to diverge from that of historical vaccinology. A good example of this is the role of the antigen ‘depot’ in establishing protective humoral immunity. Achieved through the formulation of the antigen in an emulsion or precipitate, the depot slowly releases antigen and inflammation from the injection site and into the draining secondary lymphoid tissues. Though an antigen depot has long been known to be unnecessary for achieving protective humoral immunity (Holt, 1949), it is clear that exposure of draining lymphoid tissue to antigen over many weeks does enhance the magnitude and affinity of neutralizing antibody (10–14). However, it is also equally clear that that the use of tissue-persisting antigen depots does irreparable damage to the formation of T cell memory and subsequent protective immunity (15–17). Thus, a better understanding of vaccine-elicited T cell immunity cannot always borrow from the historical rules of classic vaccinology.

Given the central role for IL-27 in mediating the T cell response to subunit vaccination, we explored the hypothesis that IL-27 might be an innate correlate of vaccine-elicited T cell immunity. Specifically, we sought to determine whether the amount of IL-27 induced by a specific adjuvant, within the lymphoid tissue draining a peripheral injection site, was predictive of the ensuing T cell response. IL-27 is a pleotropic cytokine of the IL-12 family (18), comprised of a heterodimer of EBI3 and p28 (19). As p28 is the more dynamically regulated sub-component, we generated a construct containing eGFP downstream of ~7KB of the p28 locus 5’ to the transcriptional start site (20). This transgene was used to produce a mouse (IL-27p28-eGFP) in which the GFP faithfully reports IL-27p28 expression within various APC subsets (20).

Using this IL-27p28-eGFP, we show here that p28-eGFP is induced in draining lymph node cDC1s in response to a variety of adjuvants. The amount of p28-eGFP induced by each adjuvant at 12 hours after subcutaneous vaccination successfully predicted the magnitude of antigen specific CD8 memory precursor effector cells (MPECs) as well as the magnitude of durable CD8+ T cell memory and its capacity for host protection. The strength of this correlation was such that the p28-eGFP expression could even be used to accurately predict the magnitude of the T cell response to previously untested adjuvants, indicating its potential for adjuvant screening efforts. To the best of our knowledge, this is the only instance in which a memory CD8+ T cell response can be predicted and stratified by a single innate factor produced as early as 12 hours after vaccination. Adjuvants that induce more sustained inflammation, like CpG, increased the overall frequency of antigen specific CD8+ T cells through the generation of short-lived effector cells (SLECs) without affecting the predictive capacity of the p28-eGFP/MPEC correlation. These data suggest that the frequency of memory precursors is dictated very early after vaccination and that ongoing inflammation helps form terminal effectors without diversion from the memory precursor pool.

Materials and methods

Mice

C57BL/6 mice were obtained from The Jackson Laboratory or produced in our own facility. IL-27p28-eGFP mice (20) were produced and maintained in our own facility at the University of Colorado Anschutz Medical Campus. CCR7-GFP mice (C57BL/6-Ccr7^tm1.1Dnc/J) were obtained from The Jackson Laboratory, and bred in our facility to C57BL/6 to obtain heterozygous CCR7-GFP reporter mice. All experiments were conducted in accordance with protocols approved by the University of Colorado Denver Institutional Animal Care and Use Committee.

Immunization and infection

For footpad immunization, mice were anesthetized with Isoflurane, after which 20–25 μg of detoxified whole chicken OVA (Sigma) or OVA Alexa Fluor 647 conjugate (Thermo Fisher) combined with 0.002–20 μg polyinosinic-polycytidylic acid (polyIC; GE) or 10 μg lipoteichoic acid (LTA; InvivoGen), monophosphoryl lipid A (MPL; InvivoGen), CpG (InvivoGen), 3M-012 (3M), Pam3Cys (Pam; InvivoGen), flagellin (Flag; InvivoGen), or α-Gal-Cer (aGC; Alexis Biochemicals) or a 1/10 dilution of Adjuplex (21), in a total volume of 30–40 μl PBS was administered subcutaneously in the hind footpad of the mouse. For intramuscular immunization, 20 μg OVA combined with 10 μg of polyIC in a total volume of 40ul PBS was administered to the quadriceps of anesthetized mice. For ear immunization, 20 μg OVA combined with 10 μg of polyIC in a total volume of 20ul PBS was administered between the layers of the ear of anesthetized mice. For experiments in which multiple time points were examined, immunizations were administered such that all harvests occurred simultaneously. For LM-OVA protection assays, mice that had received footpad immunizations as above 50–80 days prior received 100,000 CFU of erythromycin-resistant Lm-ova via tail vein injections. At time of infectious challenge, mice were bled via the tail vein to examine resting-state levels of memory T cells. 4 days after challenge, livers and spleens were harvested, weighed, homogenized, and subjected to cell lysis and the resulting slurry plated in serial dilution on brain-heart infusion agar plates containing 5 μg/ml erythromycin. After 36 hours of incubation at 37C, the plates were removed and resulting colonies quantified.

Isolation of cells

For isolation of dendritic and other innate immune cells from draining lymph nodes, footpad immunized mice were first euthanized by CO₂ at indicated time points. The draining popliteal lymph node was then removed and minced with tweezers in 1ml Clicks media (Irvine Scientific) containing 1 mg/ml Collagenase D (Roche) and 50 μg/ml DNase I (Worthington) per lymph node. After 45 minutes incubation at 37C, 1 ml 0.1 M EDTA was added and cells allowed to incubate another 5 minutes. Disassociated cells were then washed with HBSS (Gibco) containing 5 mM EDTA and forced through 100 μm strainers to generate single cell suspensions. For isolation of T cells from the spleen and lymph nodes of immunized mice, organs were harvested at indicated time points and forced through 100 μm strainers to generate single cell suspensions. In the case of T cells, cell suspensions were counted on a Vi-cell automated cell counter (Beckman Coulter) to determine total viable cell number. To quantify innate cells, the entire sample, representing the entire lymph node, was collected during flow cytometry analysis.

Flow cytometry

Kb-SIINFEKL tetramer was provided by the NIH tetramer core. For antibody staining, the following antibodies were used: anti-CD64 (X54–5/7.1), anti-MHCII (M5/114.15.2), BV421 anti-mouse Ly6C (HK1.4;BioLegend), BV510 anti-mouse Ly6G (1A8; BioLegend), anti-CD44 (IM7), anti-CD11c (N418), anti-XCR1 (ZET), anti-CD8 (53–6.7), anti-KLRG1 (2F1/KLRG1) and anti-CD103 (2E7) from BioLegend; anti-CD62L (MEL-14), anti-CD3 (17A2), anti-B220 (RA3–6B2), anti-CD19 (1D3), anti-CD11b (M1/70) and anti-CD127 (A7R34) from Tonbo. Samples were collected on a CytoFlex S (Beckman Coulter) and analysis performed using FlowJo software (version 10.5.3). Specific cell populations were gated on as follows: T cells- Live, B220⁻, CD11c⁻, CD3⁺; B cells- Live, B220⁺, CD11c⁻, CD3⁻; cDC1- CD3/B220⁻, CD11c⁺, XCR1⁺; cDC2- CD3/B220⁻,CD11c⁺, CD11b⁺, XCR1⁻); monocytes- CD64⁺,CD11b^hi,Ly6C^hi,Ly6G^lo.

Statistics

Prism (v8; GraphPad) was used to perform all statistical analysis. One-way ANOVAs with multiple comparisons were used to identify adjuvants which had a significantly different effect on measured outcome relative to other adjuvants. Unpaired students t-tests were used to identify differences between individual groups. Experiments were performed two or more times each, with 3 or more mice per group. Where data is correlated with other data, correlations were checked against repeats. Data indicate means ± SEM unless otherwise noted.

Results

cDC1 DCs are the primary producers of IL-27p28 following immunization

We previously published on the use of an IL-27p28-eGFP reporter mouse in which a GFP signal indicates the presence and amount of p28 expression (20). Those experiments not only validated the concurrence of the reporter with IL-27p28, but also identified i) cDC1s and monocytes as the primary producers of IL-27p28 in the spleen after IV adjuvant administration, and ii) a direct correlation between cDC1 reporter expression at 4 hours with CD8+ T cell expansion at 7 days in the spleen post IV immunization (20). Because route of immunization can influence the activation of innate immunity and the subsequent magnitude, duration, and localization of the T cell response to immunization, we examined the p28-eGFP reporter expression in APC subsets within the draining lymph node after various forms of local vaccine administration. In order to minimize any potentially confounding influences of the antigen (tissue retention/bioavailability, lymphatic draining, antigen uptake, class I binding, etc.), we utilized the model antigen ovalbumin (OVA), for which numerous cellular and molecular tools are available for dissecting immune response parameters and mechanisms. Initially, mice were immunized subcutaneously in the footpad with OVA and polyIC. At 12 hours post immunization, the draining popliteal lymph nodes (dPLN) were harvested, digested into a single cell suspension, and the p28-eGFP reporter expression determined in T cells, B cells, cDC1 (CD3/B220⁻, CD11c⁺, XCR1⁺), cDC2 (CD3/B220⁻, CD11c⁺, CD11b⁺, XCR1⁻), and monocytes (CD64⁺,CD11b^hi,Ly6C^hi,Ly6G^lo) by flow cytometry. Consistent with previous reports (20, 22, 23), cDC1s were the primary producers of IL-27p28 (Fig. 1, A and B), with 40–70% of cDC1s in the dPLN p28-eGFP+ following immunization. cDC2s and monocytes also expressed the p28-eGFP reporter (Fig. 1 B), though at significantly lower frequencies than cDC1s. Based on a correlation we previously established between IL27p28 message and the p28-eGFP reporter geometric mean fluorescence intensity (gMFI), we noted that cDC1s also produce substantially more IL-27p28 on a per cell basis (Fig. 1 C). cDC1s also produced IL-27p28 within draining lymph nodes after either intradermal (ear) or intramuscular (IM) immunization (Fig. 1, D and E), indicating that this was not a phenomenon exclusively associated with the footpad/dPLN.

Figure 1:

IL-27p28 is made by cDC1s in the draining node following peripheral subunit immunization. (A-E) IL-27p28-eGFP mice were immunized in the footpad (A-C) or footpad, ear or thigh muscle (D, E) with 10ug polyIC and 25ug OVA, and dLNs harvested 6–12 hours later. (A) Representative gating strategy identifying cDC1s and demonstrating GFP signal following immunization. (B, C) cDC1s, cDC2s, monocytes, B cells and T cells were identified in order to determine IL-27p28-eGFP production by each cell type. (D, E) the GFP signal from cDC1s was measured in the cervical, popliteal and inguinal dLNs following immunization between the ear layers, in the footpad or in the thigh respectively. Data indicate means ± SEM, n ≥ 3 mice per group, representative of 2 or more experiments. **p < 0.01, ****p < 0.0001.

Given the short lifespan of DCs (24, 25) and the corresponding narrow window for T cell activation, we next examined the timing of cDC1 activation, antigen uptake, and p28-eGFP reporter expression in the dPLN following footpad immunization. Initially, B6 mice were immunized with polyIC, dPLNs were harvested at various points following immunization, and the relative frequencies of cDC1 and cDC2 were analyzed. We observed two peaks in overall DC numbers, at 12 and 24 hours. In both frequency (Fig. 2 A) and total numbers (Fig. 2 B), representation of the two DC subsets shifted from predominantly cDC2s at 0 hours to majority cDC1s at 12 hours, to being split between cDC1s and cDC2s by 24 hours. Over this time, DCs also increased in class II expression (Fig. 2 A) and in number (Fig. 2 B), consistent with their activation and accumulation. DC accumulation within the node can occur through proliferation of DC precursors (24–26) and accumulation of migratory DCs (27, 28). As previous studies have shown numbers of migratory DCs peaking at ~24 hours (29), our data suggested the possibility of an early (12 hour) peak in resident DC expansion/frequency and a subsequent (24 hour) peak in migratory DCs. We confirmed this hypothesis using heterozygous CCR7-GFP knock-in reporter mouse, in which essentially all DCs harvested at 12 hours post immunization were resident (CCR7 low) (Fig. 2, C and D), while by 24 hours, the DC pool was split between resident and migratory (CCR7 high) DCs, with a decline in the number of resident cDC1s (Fig. 2, C and D). However, total (resident + migratory) cDC1 frequency and numbers were generally highest at 12 hours post immunization (Fig. 2, E and F). Using the p28-eGFP reporter host to assess IL-27p28 production, we observed that the number of IL-27p28 producing cDC1s was also highest at 12 hours (Fig. 2 G). The gMFI of the p28-eGFP did not change substantially in cDC1s between 12 and 24 hours (Fig. 2 H), indicating that the amount of IL-27p28 produced on a per cell basis is not substantially different between migratory and resident cDC1s. Together, these results indicate that following peripheral immunization, i) both resident and migratory cDC1s are the primary producers of IL-27p28 in the draining node, ii) the number of IL-27p28-producing cDC1s peaks at 12 hours post immunization, and iii) this peak consists primarily of resident cDC1s.

Figure 2:

cDC1 IL-27p28 production in the dPLN peaks at 12 hours after immunization. (A-H) C57BL/6 (A, B), C57BL/6-Ccr7^tm1.1Dnc/J (C, D) or IL-27p28-eGFP (E-H) mice were immunized in the footpad with 10ug polyIC and 25ug OVA and dPLNs harvested and examined by flow cytometry at the indicated time point. (A, B) Gated on total DCs. (C-H) Gated on total DCs or as indicated. Data indicate means ± SEM, n ≥ 3 mice per group, representative of 2 or more experiments. **p < 0.01, ****p < 0.0001.

Early LN IL-27p28 predicts the frequency of circulating antigen specific T cells

We next examined the cDC1 response at 12 and 24 hours after footpad immunization to a range of adjuvants. At 12 hours, polyIC had the greatest impact on percent p28-eGFP+ cDC1s (Fig. 3 A), the total number of p28-eGFP+ cDC1s (Fig. 3 B), and cDC1 p28-eGFP gMFI (Fig. 3 C). PolyIC cDC1s continued to show the highest percent p28-eGFP+ and p28-eGFP gMFI at 24 hours (Fig. 3, A and C), but by this time point CpG was greatest in regard to total cellularity (Supplemental Fig. 1 A), total DCs (Supplemental Fig. 1 B), total cDC1s (Supplemental Fig. 1 C), and p28-eGFP+cDC1s (Fig. 3 B), with polyIC matched in all these values by the TLR7 agonist 3M-012.

Figure 3:

cDC1 IL-27p28 production at 12 hours in the dPLN correlates with systemic T cell counts at 7 days. (A-C) IL-27p28-eGFP mice were immunized in the footpad with 10ug indicated adjuvant and 25ug OVA and dPLNs harvested 12 and 24 hours later. Gated on cDC1s. (D, E) C57BL/6 mice were immunized in the footpad with 10ug indicated adjuvant and 25ug OVA and dPLNs, blood and spleens were harvested 7 days later. Gated on CD8s. (F) Linear regressions of correlations between data in Fig. 3 C and Fig. 3 E. Data indicate means ± SEM, n ≥ 3 mice per group, representative of 2 or more experiments.

We next established the magnitude of the CD8 T cell response generated by each adjuvant. At 7 days following footpad immunization, dPLNs, blood and spleens were harvested, and the magnitude of the T cell response was examined by tetramer staining (Fig. 3 D). CpG produced by far the largest T cell response within the dPLN, followed by all others at a much lower level (Fig. 3, D and E). In contrast to the response in the node, a different hierarchy of adjuvant efficacy was seen in the number of circulating antigen specific T cells in the spleen, in which polyIC produced the greatest response, followed by MPL and then the remaining adjuvants (Fig. 3, D and E, lower panels). This hierarchy also held true for the blood (Supplemental Fig. 1 D), indicating that the number of antigen specific T cells still in the draining node seven days after subunit immunization does not correlate with the number of circulating adjuvant-elicited T cells (Supplemental Fig. 1 E). This divergence in dPLN vs systemic T cell numbers tracks somewhat with differences in sustained draining node inflammation (as measured by DC and total LN cellularity) (Supplemental Fig. 1, A and B), and suggests that the numbers of T cells in the draining LN late after immunization may better reflect LN retention of T cells than the efficacy of their generation.

We next assessed whether any particular cDC1 parameter from the dPLN was predictive of the splenic CD8 T cell response. We compared the various early response metrics described above with the corresponding number of T cells in the spleen resulting from each adjuvant. We found that total cDC1 numbers (Supplemental Fig. 2, A and B) and even p28-eGFP cDC1 numbers (Supplemental Fig. 2, C and D) at 12 or 24 hours post immunization were not predictive of T cell numbers 7 days later. In contrast, we found a strong correlation between the cDC1 p28-eGFP gMFI in the draining node at 12 hours and T cell numbers in the spleen, but not the dPLN (Fig. 3 F). In keeping with our previous data (20), this correlation held best with the p28-eGFP gMFI, but not the number of cells producing IL-27p28. The strength and significance of this correlation held even when eliminating polyIC from consideration (Supplemental Fig. 2 E), indicating that the values associated with the polyIC response are not unduly influencing the predictive value of the 12 hour p28-eGFP gMFI signal. Interestingly, the significance of the correlation between T cell response and p28-eGFP gMFI did not hold for the cDC1s at 24 hours (Supplemental Fig. 2 F).

IL-27p28 predicts adjuvant-elicited CD8+ T cell memory precursors and long-term memory

Ultimately, the goal of vaccination is the generation of a stable pool of protective memory cells. This pool is heavily influenced by the peak magnitude of the primary response, a time point that can vary across different adjuvants for a variety of reasons. Indeed, though the peak response to adjuvants such as PolyIC and MPL are established at 7 days post-immunization, the response to CpG often peaks between day 7 and day 10 (Supplemental Fig. 3 A). In quantifying the total T cell response at this later time point, we observed a loss of correlation with the 12 hour LN cDC1 p28-eGFP gMFI (Supplemental Fig. 3 B), the result of an expansion of CD127^loKLRG^+/− T cells (Fig. 4, A and B). However, the predictive power of the p28-eGFP gMFI held with respect to the number of MPEC T cells generated by each adjuvant (Fig. 4 C). The number of SLECs and MPECs did not correlate with one another (Supplemental Fig. 3 C), with CpG appearing to drive a large expansion of SLECs with minimal impact to the MPEC pool (Fig. 4 B).

Figure 4:

cDC1 IL-27p28 production at 12 hours correlates with protection against infection at memory time points. (A, B) C57BL/6 mice were immunized in the footpad with 10ug indicated adjuvant and 25ug OVA and dPLNs, blood and spleens were harvested 10 days later (data from spleen shown). (C) Linear regression of correlation between data in Fig. 3 C and Fig. 4 B. (D) C57BL/6 mice were immunized in the footpad with 10ug indicated adjuvant and 25ug OVA and spleens were harvested 50–80 days later. (E) Linear regression of correlation between data in Fig. 3 C and Fig. 4 D. (F) Memory stage mice from Fig. 4 D were challenged intravenously with 1.5×10⁵ LM, and 4 days later livers were harvested, homogenized and plated in serial dilution on BHI plates containing erythromycin. CFU of LM obtained per gram of liver. (G) Nonlinear regression of correlation between data in Fig. 3 C and Fig. 4 F. Data indicate means ± SEM, n ≥ 3 mice per group, representative of 2 or more experiments.

To determine if the trend in MPEC numbers was related to memory, we next examined the T cell memory pool at 80 days after immunization and its correlation with 12 hour LN IL-27p28 production. Mice immunized with adjuvants that gave rise to large pools of SLECs at early time points also contained relatively large reservoirs of effector-like (KLRG1⁺, CD127⁻) cells at memory time points (Fig. 4, D). This meant that, as with 10 days following immunization, IL-27p28 production at 12 hours was not predictive of total frequency of antigen specific T cells at these late time points (Supplemental Fig. 3 D). However, 12 hour cDC1 p28-eGFP gMFI continued to be highly predictive of the pool of KLRG⁻, CD127⁺ cells (Fig. 4, E). Given the demonstrated protective capacity of CD127^hi T cells (30), we predicted that the frequency of these memory cells after immunization would also correlate with protection against subsequent infection. To this end, we challenged memory-stage mice with ovalbumin-expressing Listeria monocytogenes (LM-OVA) and measured the bacterial load in the liver after 4 days as an indicator of CD8+ T cell-mediated host protection. As predicted, we found a strong correlation between the frequency of CD127^hi cells and control of infection 80 days later (Fig. 4, E and F). Because of the association between early IL-27p28 and memory cell frequency (Fig. 4 E), this meant that IL-27p28 production at 12 hours post immunization successfully predicted T cell-mediated host protection 80 days later (Fig. 4 G).

As noted previously, the strength of the IL-27p28:T cell correlation was not simply due to the signal derived from the response to polyIC, as a statistically significant correlation was maintained even after eliminating polyIC from consideration (Supplemental Figs. 2 E and 3 E). To ensure that our correlation was not coincidental with the amount of adjuvant utilized, we immunized mice with a dose titration of polyIC and again tracked 12 hour cDC1 p28-eGFP gMFI and T cell generation. Utilizing 10-fold titrations, we observed a peak of p28-eGFP gMFI at 2ug of polyIC injected into the footpad, which consistently declined slightly at 20ug (Fig. 5 A), a common characteristic of the response to innate receptor agonists (31). As predicted by our previous data, the magnitude of the CD8 T cell response (Fig. 5 B) continued to correlate with the amount of IL-27p28 produced by cDC1s at 12 hours (Fig. 5 C).

Figure 5:

Early IL-27p28 production predicts the efficacy of novel adjuvants in generating protective T cell responses. (A) IL-27p28-eGFP mice were immunized in the footpad with indicated amount of polyIC and 25ug OVA and dPLNs harvested 12 later. (B) C57BL/6 mice were immunized in the footpad with indicated amount of polyIC and 25ug OVA and spleens were harvested 10 days later. (C) Linear regression of correlation between data in Fig. 5 A and Fig. 5 B. (D) IL-27p28-eGFP mice were immunized in the footpad with 10ug adjuvant and 25ug OVA and dPLNs harvested 12 hours later. (E) gMFIs for previously unexamined adjuvants from Fig. 5 D were normalized against gMFIs obtained from polyIC-immunized mice harvested and analyzed simultaneously, and these adjusted gMFI levels were used to calculate predicted T cell numbers (grey boxes) using the linear regression formula from Fig. 4 C. Experimental values (black circles) were obtained by immunizing C57BL/6 mice in the footpad with 10ug indicated adjuvant and 25ug OVA and harvesting spleens 10 days later. (F) Overlay of values obtained in Fig. 5 D and 5 E with correlation plot from Fig. 4 C. Data indicate means ± SEM, n ≥ 3 mice per group, representative of 2 or more experiments.

As the data overwhelmingly supports a predictive relationship between IL-27p28 and CD8+ T cell MPEC and memory formation, we next used the existing model to predict the T cell response to untested adjuvants. We selected two adjuvants, the TLR5 agonist Flagellin (32) and the CD1-binding NKT cell stimulus αGalCer (αGC) (33), and measured the p28-eGFP gMFI of p28-producing cDC1s 12 hours after footpad immunization (Fig. 5 D). Using the polyIC-induced cDC1 GFP gMFI in this experiment, we normalized the p28-eGFP gMFI values for each adjuvant to the mean polyIC p28-eGFP gMFI in the experiment from Fig. 4 C. We then used the associated line equation (in which y is the normalized gMFI and x is the T cell value) to obtain a predicted range of MPECs for each adjuvant (Fig. 5 E, shaded boxes). This predicted range fit well within the standard deviation of the actual number of MPECs, obtained from B6 mice immunized concomitant with the p28-eGFP mice and analyzed 10 days post immunization (Fig. 5 E, closed circles). When overlaid onto the correlation from Fig. 4 C, the error of the observed values fell within the previously generated linear regression, indicating a high level of predictability (Fig. 5 F). Thus, IL-27p28 production from cDC1s at 12 hours accurately predicts the efficacy by which an adjuvant will generate CD8 T cell memory.

Discussion

Previously we published that following systemic immunization, IL-27 production in the splenic cDC1 subset could predict the magnitude of the bulk T cell response at 7 days (20). Missing from this previous data, however, were the use of a route more appropriate to typical vaccinations or any correlates to long term memory responses. The present data address these important issues in vaccine biology, utilizing the p28-eGFP reporter to stratify commonly utilized vaccine adjuvants and predict consequent memory T cell frequency and protective capacity. Though IL-27’s influence on T cell memory has been noted, the literature is contradictory as to whether this relationship is affirming (3, 34), or opposing (35, 36). Unique to the data presented here, however, is our conclusion of the unexpected and unprecedented predictive value that the amount of IL-27p28 produced per cDC1 has on the magnitude of CD8+ T cell memory.

In addition to pairing with EBI3 to form IL-27, IL-27p28 may also be secreted alone as a distinct cytokine known as IL-30 (37–39). As our reporter tracks only p28 expression, our results could conceivably identify a robust correlation between CD8+ T cell memory formation and cDC1 IL-30, IL-27 or some combination of both. That being said, two lines of evidence point toward IL-27 as the more relevant cytokine. First, as mentioned above, the T cell response to subunit immunization is reliant on IL-27 signaling through T cell-expressed IL-27R (3). That is, the role of IL-27 in vaccine elicited immunity can be observed under conditions of T cell-specific IL-27R deficiency. While the specific receptor for IL-30 is as of yet undefined, it does not appear to be the IL-27R. Second, using an ELISA with capture specificity for IL-27p28 and detection of EBI3 (that is, whole IL-27), high amounts of IL-27 are detected in the serum after vaccination and in the same timeframe as cDC1 p28-eGFP expression (Supplemental Fig. 3 F). These observations indicate that it is most likely IL-27, not IL-30, that correlates with and predicts consequent CD8+ T cell memory formation.

It is interesting that while some adjuvants augment the expansion of terminally differentiated effectors, this expansion does not seem to subtract from the MPEC/memory cell pool (Figure 4). These conclusions are consistent with the model proposed by Chang and Reiner (40) whereby asymmetric segregation of fate-determining signaling and transcription factors generates daughter cells with different cell fates, ensuring the production of effector cells without compromising memory cell frequency. Additionally, it appears that the production of terminal effectors occurs after the initial immunization and T cell activation phase, perhaps related to the persisting inflammatory properties of the adjuvant utilized (41, 42). Though the total frequency of antigen specific T cells is augmented by these adjuvants, this has no additional protective value against infectious challenge, perhaps related to the demonstrated negative consequences of a persisting antigen depot (15–17).

Though both resident and migratory cDC1s produce IL-27p28 in the draining node following peripheral immunization, only resident DC IL-27p28 production is predictive of the T cell response since this is the only population meaningfully present at 12 hours post immunization. This was surprising given previous data by Jenkins and colleagues showing the importance of the migratory DCs in maximizing the CD4 response to intradermal vaccination in the ear (29). It seems further distinct from the results of Heath and colleagues in which migratory DCs were critical for simulating naïve T cells responding to HSV-1 whereas resident DCs were necessary for stimulating memory T cells (43). However, our data do not negate an important role for migratory cDC1s in our vaccine-elicited responses, they merely indicate that migratory cDC1 production of p28 is less predictive of downstream immunity than that produced by LN resident cDC1s. Whether or not there is biology underlying this difference in predictive capacity specific to IL-27 signaling remains to be determined.

The magnitude of an immune response from local (subcutaneous, intradermal, or intramuscular) vaccine administration is often measured in the draining LN at the presumed peak (7–10 days) of the immune response. Our data indicate, however, that T cell number in the draining node at this time point correlates poorly with the overall magnitude of the T cell responses systemically. Given that CD8+ T cells begin to migrate out from their initial site within a few days of antigen encounter, this is perhaps unsurprising, but it highlights the fact that caution should be used when using LN T cell numbers as a predictor of vaccine efficacy. Needing much more exploration is the role of activated T cell retention within the node, its relationship to the type of adjuvant utilized, and whether this retention within the node is beneficial or detrimental to downstream immunity.

It is intriguing to note that it was the cDC1 p28-eGFP gMFI, not %GFP+, that best predicted downstream MPEC and memory formation. Thus, the amount of IL-27 produced per cDC1 appears to set the extent to which memory precursors, and ultimately memory T cells, are formed. This is at least consistent with the hypothesis that a T cell preferentially interacts with p28-producing cDC1s, and that this interaction is a determining factor in the magnitude of the ensuing response. Experiments utilizing multiphoton microscopy are currently investigating the propensity cDC1 IL-27 production to influence the quantity/quality of T cell:DC interactions.

Finally, results from the titration of polyIC indicate that IL-27p28 production is not only a good biomarker for adjuvant efficacy, but also for optimizing adjuvant dosing levels. Though much thought is often given to the appropriate dosing of antigen, less seems directed toward identifying the proper dosing of adjuvant, despite the growing recognition of the deleterious impact of excessive inflammation on tissue homeostasis and long-lived immune memory. Both CD4 and CD8+ T cell responses are IL-27 dependent (3), so while we have only documented a predictive relationship between cDC1 expression of IL-27 and CD8+ T cell elicitation, it seems likely that this will hold true for CD4+ T cells as well. It is as of yet unclear to what degree IL-27 might predict humoral immunity. Since the majority of vaccines are directed toward the generation of antibodies, it will be important to understand the role of IL-27 in B cell activation, and whether any predictive correlations can be extracted from its interplay with IL-27-dependent CD4+ T cell responses.

Key Points.

• 12 hour cDC1 IL-27p28 expression predicts adjuvant elicited CD8 T cell memory.

• cDC1 IL-27p28 expression uniquely stratifies T cell-inducing adjuvants.