The human study was reviewed and approved by the Medical Ethical Committee Brabant (Tilburg, Netherlands, NL51305.028.15) and all donors provided written informed consent.

The human study cohort has been previously described (28). In brief, Q-fever exposed individuals were recruited from a cohort characterized in a previous large Q fever study conducted in the village of Herpen, the Netherlands (29), which experienced a high incidence of C. burnetii infection during the 2007-2010 Q fever outbreak (30). In this previous study, 80% of the adult village population was screened in 2014 for evidence of adaptive immunity towards C. burnetii. Individuals were invited to participate in this current study following pre-selection based on clinical history and adaptive C. burnetii-specific immune responses determined in 2014. All study participants, including immunologically C. burnetii-naive controls, resided during and following the Q fever outbreak in the same area and were assigned to study groups based on clinical history and immunological assays performed at the time of study enrollment in 2015. Cellular immunity to C. burnetii was assessed with the CE-marked Q-Detect interferon-γ release assay (31) as described previously (28). To maximize the potential to detect C. burnetii-specific T cells in this cohort who were predominantly enrolled for epitope-screening (28), preference was given to donors with strong responses to whole heat-killed C. burnetii in the IGRA and without potentially confounding immune disorders. In total, 143 participants provided written informed consent. IGRA responses were re-assessed upon enrollment in October 2015, and serological status was additionally assessed by immunoblot. Volunteers were allocated to control Group 1 if they had no history of Q fever disease (29), were negative (32) by IGRA and by immunofluorescence assay (IFA) in spring 2014, and were again negative by IFA and immunoblot when enrolled in the previous study (28) in autumn 2015. The remaining volunteers that were positive by IGRA in spring 2014 were subdivided based on their serostatus (IFA in spring 2014 and immunoblot analysis in autumn 2015) and past Q fever disease (either registered (notified) in the national surveillance system, or self-reported) into Group 2 (seronegative, asymptomatic), Group 3 (seropositive, asymptomatic) and Group 5 (seropositive, symptomatic). A single volunteer fell into Group 4 (seronegative, symptomatic) and was not included in further analysis. At the time of enrolment, all volunteers were convalescent from their original exposure to C. burnetii, and none of the volunteers included in this study had known active C. burnetii infection or were diagnosed as suffering from acute or chronic Q fever.

All peripheral blood samples for this study were obtained between July 2016 and May 2017, i.e., six to ten years after initial exposure during the 2007-2010 Q fever outbreak in the Netherlands. A total of n=64 individuals contributed to one or more aspects of this immune profiling study ( Tables 1 , 2 ).

The Q-Detect™ assay has previously been described in detail (31). In brief, whole lithium-heparin anti-coagulated blood was stimulated with C. burnetii antigen (heat killed Cb02629, a strain isolated during the Dutch Q fever outbreak, lot 14VRIM014 prepared by Wageningen Bioveterinary Research from a master cell bank using a cell-free culture method and quality controlled for protein concentration, functional TLR stimulation, and stimulation of IFNγ in samples from known Q fever-exposed individuals). Assays were performed in 96-well polypropylene plates (Greiner BioOne) by adding 180 µl blood to 20 µl C. burnetii antigen diluted in phenol red-free RPMI supplemented with Glutamax (2 mM), Gentamycin (5 µg/ml) and sodium pyruvate (1 mM, all ThermoFisher Scientific). A 1.5% (v/v, final concentration) solution of PHA-M (ThermoFisher Scientific) was added to separate wells for each sample as a positive control. Medium only was added to the negative control wells for each sample, results from which were used to correct for any background levels of IFNγ in the sample. All stimulations were performed in duplicate. After 22 ± 1 hours whole blood cultures were re-suspended. IFNγ concentrations were assessed in whole blood by ELISA using the IFNγ Pelipair protocol (Sanquin) with minor modifications. The upper detection limit of IGRA under these conditions is 1050 pg/ml. A subject was scored positive by IGRA if the C. burnetii antigen induced IFNγ production was ≥16 pg/ml above background and the ratio of the logarithmic value of background-subtracted C. burnetii antigen and PHA responses ((log[C. burnetii]-log[neg control])/(log[PHA]-log[neg control])) was ≥0.4.

Data were retrieved from the Helios Mass Cytometer into FCS files. EQ calibration beads were used to normalize all FCS files, to minimize any variation that occurred during data acquisition on the Helios Mass Cytometer. This normalization was performed using the Fluidigm software and sample data were debarcoded using the Fluidigm Debarcoding software. Post-processing, following identification of viable singlet CD45⁺CD66b^- mononuclear cells, four major immune populations (CD4 T cells, CD8 T cells, B cells and innate immune cells) were manually gated using FlowJo 10 ( Supplementary Figure S3 ).

Several markers showed strong run-to-run or day-to-day staining variation. For manual gating of CD14⁺ monocytes and cytokine producing cells, gates were therefore set per run based on the run-specific staining control sample. Prior to unsupervised clustering analysis, marker expression levels had to be additionally normalized across the four runs. This normalization, based on representative concatenate samples (see below), was performed separately for each immune subpopulation using CytoNorm (33) to remove batch-specific variations. Of the 37 antibodies included in the CyTOF panel, four markers (CD45RO, CD45RA, CD197/CCR7, CD19) showed very large batch-to-batch variation outside the range for effective normalization and were excluded from unsupervised clustering analysis. The canonical markers used to identify the four main cell populations (CD45, CD66b, CD3, CD4, CD8, CD20) were not included into the normalization process. The following mass cytometry markers were normalized using CytoNorm: CD16, FOXP3, CD56, CD33, IL-2, CD137, CD154, CD14, IL-6, IFNγ, T-bet, IL-8, HLA-DR, CD69, CD206, CD70, CD27, IL-4, TNFα, IL-1β, CD11c, IL-17, GATA3, CD38, CXCR4, TGFβ, and IL-10. A representative normalization sample was created for each of the four runs by down-sampling each debarcoded sample and concatenating the data. The four representative concatenate samples created in this manner were then used as the normalization reference samples in CytoNorm, to aid normalization of the data across all 4 runs. Since normalization was performed separately for each immune subpopulation, normalized expression levels for a given marker cannot be compared across the major immune populations. To assess if the normalization was accurate, we compared the control unstimulated PBMC sample included in all four runs as a staining control, before and after normalization ( Supplementary Figures S4–S11 ). UMAP dimension reduction was performed on the four control samples prior to normalization, and separately after normalization. Before normalization, the surface and intracellular marker data for this control sample from different runs behaved like different samples. Post-normalization the control sample from each run overlaid the others, indicating that they were properly normalized.

Following normalization, data for all samples from all runs were clustered for each major cell population separately. Clustering was performed using FlowSOM (34), which employs self-organizing maps to identify immune cell clusters. Four major parental cell populations were defined as follows: CD4 T cells (CD3⁺CD4⁺CD8^-CD20^-); CD8 T cells (CD3⁺CD4^-CD8⁺CD20^-); B cells (CD3^-CD20⁺); innate cells (CD3^-CD20^-). The mass cytometry markers used to cluster CD4 T cells, CD8 T cells, B cells and innate cells are listed in Table 6 . FlowSOM parameters were initially set to identify 900 clusters for each major cell population. These 900 were subsequently meta-clustered to identify 20 meta-clusters from each major cell population (C1-C20). Using the SpadVizR R package, the median intensity of each marker was plotted in each meta-cluster to enable the phenotypic definition of the cell population via parallel coordinate plots. For each donor, the frequencies of each meta-cluster within each parental population in both control and C. burnetii antigen-stimulated conditions were calculated as a percent of the total number of cells clustered from that donor for that cell population. Subsequently, background corrected frequencies for each meta-cluster were calculated to identify whether the respective meta-cluster increased or decreased upon stimulation with C. burnetii antigen.

Statistical analysis was conducted based on the original group assignments (Groups 1, 2, 3, and 5), as well as combined group assignments, such as Group 2 + 3 + 5 (IGRA-positive). Statistical analysis was performed using GraphPad Prism 9. Spearman’s correlation analysis was conducted for all clusters and plasma cytokine secretion levels using R 4.1 and Hmisc. Heatmaps and dendrograms were created using Python 3.10 using Pandas, Pyplot and Seaborn. tSNEs were generated using Python 3.10 and Scikit-Learn.

The individuals included in this study were divided into four groups: Those with no apparent prior exposure based on the absence of humoral and cellular adaptive response (negative IFA and IGRA) and lack of clinical history (Group 1), and previously exposed individuals based on a measurable cellular response by IGRA. The presence of a cellular immune response to C. burnetii antigen suggests that the IGRA-positive individuals had experienced prior infection with C. burnetii even if the infection did not result in a clinical history of symptomatic Q fever. These IGRA-positive individuals were further subdivided based on serological and prior disease status into Group 2 (seronegative, asymptomatic), Group 3 (seropositive, asymptomatic) and Group 5 (seropositive, symptomatic). No individuals with known active C. burnetii infection, as either current acute or chronic Q fever, were included in the study.

We first assessed whether the exposure status of individuals was associated with differential release of cytokines previously reported to be associated with re-stimulation responses to whole-cell C. burnetii antigen. We compared bulk cytokine responses as determined by multiplex V-PLEX assay following whole blood stimulation with heat-killed C. burnetii antigen in individuals with different states of prior exposure, clinical history, and serological status. Prior exposure (Groups 2, 3 and 5) was associated with significantly elevated release of IFNγ quantified by V-PLEX (p<0.0001; Figure 1A ), consistent with the positive IGRA results used to define groups at study enrollment. IGRA responses (at enrollment) and IL-2 responses showed a strong correlation (Spearman Rho = 0.86, p= 4.5535E-09), and IGRA positivity was associated with significantly increased release of IL-2 (p>0.0001) compared to that observed in cells from unexposed individuals from group 1 ( Figure 1A ). In contrast, the innate cytokines IL-10, IL-1β and TNFα were released in response to C. burnetii antigen-stimulation by cells from both control (Group 1) and pre-exposed individuals (Groups 2, 3 and 5). Median levels of the three innate cytokines were elevated in pre-exposed individuals, although statistical analysis supported a trend toward differential levels only for IL-1β (p=0.062) ( Figure 1A ). No significant correlation was found between innate and adaptive cytokine responses, or amongst innate cytokines.

Dimensional reduction of whole blood cytokine release data from n=55 individuals by t-SNE followed by k means clustering of t-SNE embeddings identified two main clusters of study participants ( Supplementary Figure S12 ). Cluster 1 comprised all control Group 1 individuals, a majority of Group 2 and n=1 donor from Group 3 who showed low IFNγ and IL-2 responses to stimulation with C. burnetii. Cluster 2 contained all remaining IGRA-positive individuals showing predominantly high IFNγ and IL-2 responses. TNFα, IL-1β and IL-10 responses to C. burnetii antigen did not correlate with the cluster groupings.

Further analysis of the clinical subgroups defined by clinical and immunological status showed that in contrast to the IFNγ responses used to define prior exposure, IL-2 responses were significantly elevated only in seropositive individuals (Group 3 and 5) regardless of symptoms during infection, while IL-1β release was elevated specifically in seropositive individuals who had convalesced from symptomatic infection (Group 5) ( Figure 1B ). Finally, TNFα release was significantly elevated only in the small group of pre-exposed individuals with past asymptomatic infection who were analyzed (n=5, Group 3).

Except for IFNγ, released cytokine levels in seronegative Group 2 individuals were indistinguishable from those in the control Group 1 ( Figure 1B ), and median IGRA responses in Group 2 were lower than in seropositive Group 3 and 5 individuals ( Table 1 ). These lower cellular responses in individuals that fail to show sero-conversion has previously already been described in a much larger cohort study in C. burnetii-exposed individuals (31) as well as in individuals exposed to HIV (35–37), HBV (38), HCV (39), HSV-2 (40) and SARS-CoV-2 (41, 42).

Overall, of the bulk cytokine responses evaluated, IFNγ and IL-2 were the most informative for inferring prior exposure. Innate cytokines TNFα and IL-1β also showed some association with prior exposure by natural infection, consistent with previous results (43), but elevation was more marginal and restricted to specific subgroups of pre-exposed individuals.

To explore the cellular source of these cytokines in C. burnetii antigen-stimulated whole blood cultures, we first performed flow cytometry analysis for a randomly selected set of n=36 donors from the two groups with the greatest difference in clinical and immunological status: IGRA-negative and IFA-negative individuals with no clinical history of Q fever (Group 1), and IGRA-positive and IFA-positive individuals with past symptomatic infection (Group 5). We focused on the assessment of C. burnetii-induced intracellular cytokine production by monocytes and CD4 T cells, effector cells that are known to promote killing of C. burnetii and clearance of infection (44, 45).

In C. burnetii antigen-stimulated whole blood, a high proportion of monocytes from both controls (Group 1) and IGRA-positive donors with a clinical history of Q Fever (Group 5) produced TNFα (median 10.8% and 17.6%), IL-1β (median 55.8% and 66.6%) and IL-6 (median 43.3% and 55.5%). Only the difference in IL-6⁺ monocytes between Groups 1 and 5 reached statistical significance (p = ≤0.05) ( Figure 2A ). However, there was a modest trend toward higher proportions of TNFα, IL-1β and IL-6 producing monocytes in pre-exposed Group 5 individuals.

T cell responses were assessed in two flow cytometry experiments, each including n=10 donors from each of Groups 1 and 5. Initially, a single T cell panel was used to assess IFNγ and TNFα production and the activation marker CD154 ( Figure 2B ). In a second round IL-2 production was also interrogated using a second T cell panel ( Figure 2C ). CD4 T cells consistently showed significantly higher proportions of IFNγ⁺, TNFα⁺ and CD154⁺ cells ( Figures 2B, C ) as well as IL-2⁺ cells ( Figure 2C ) in Group 5 individuals with clinical history of the disease compared to Group 1 controls. CD3⁺ CD4^- T cells, inferred as CD8 T cells, showed no significant difference in C. burnetii-induced TNFα or IL-2 production or CD154 expression between control (Group 1) and convalescent (Group 5) individuals ( Figures 2B, C ). However, amongst the Group 5 individuals assessed in round 1, a small number did show CD4^- IFNγ T cell responses of a similar magnitude as CD4 T cells ( Figure 2B ). Amongst the individuals in Group 5 assessed in round 2, this IFNγ response in CD4^- T cells was statistically significant and comparable to the response of CD4 T cells ( Figure 2C ). This suggests that CD8 T cells also contribute to recall IFNγ responses.

The flow cytometry panels focused on assessing responses by monocytes and CD4 T cells. To characterize the immune cell subsets involved in recall responses to C. burnetii antigen more deeply and broadly, we utilized a highly multiplexed CyTOF panel in combination with manual gating and unsupervised clustering analysis. The CyTOF panel included additional lineage markers to investigate responses by CD8 T cells, B cells and innate immune cells, as well as a wider variety of activation markers and cytokines. For this CyTOF analysis we further expanded the selection of subjects: in addition to individuals from Group 1 (controls) and Group 5 (seropositive, symptomatic), we also included Group 2 (seronegative, asymptomatic) and Group 3 (seropositive, asymptomatic) individuals.

Cytokine responses were initially quantified in total CD45⁺CD66b^- mononuclear cells by manual gating. In C. burnetii antigen-stimulated PBMCs, the proportions of cells showing C. burnetii-specific production of IFNγ, IL-6, IL-10 and TNFα were significantly higher in pre-exposed IGRA-positive individuals compared to Group 1 controls ( Figure 3A ), consistent with IGRA responses at study enrollment ( Table 1 ). The median increase with pre-exposure in cells producing these cytokines was two-fold or less, except for the proportion of IL-6⁺ cells which increased nearly six-fold. In contrast, the proportion of cells producing the innate cytokines IL-1β and IL-8 in response to C. burnetii-stimulation was comparable across the two groups and thus independent of prior exposure status. IFNγ levels were specifically elevated in seropositive Group 3 and 5 individuals, and the same was true for IL-10 and TNFα ( Figure 3B ). Although the proportion of IL-2-producing cells was not significantly different between IGRA-negative Group 1 controls and all IGRA-positive donors (Group 2 + 3 + 5) ( Figure 3A ), seropositive Group 3 and 5 individuals showed clearly higher IL-2 responses compared to Groups 1 and 2 ( Figure 3B ), with the difference between Group 2 and 3 individuals reaching statistical significance. C. burnetii-specific IL-4 responses were not detectable, and IL-17 responses were only detectable in a very minor proportion of cells (<0.05%) and showed no differences between IGRA-negative and IGRA-positive groups (data not shown).

To assess the cellular source of these cytokines, we separately analyzed CD45⁺ mononuclear cells positive for each cytokine for individuals who were either IGRA-negative (Group 1) to IGRA-positive (Group 2 + 3 + 5) ( Figure 4 ). IFNγ was broadly produced by all immune cell types evaluated, not solely by CD4 and CD8 T cells, regardless of the status of prior exposure. However, IFNγ expression by CD4 T cells (p = 0.029) and CD56^- innate cells (presumably monocytes and dendritic cells, p = 0.005) increased with prior exposure (higher proportion in Group 2 + 3 + 5 compared to Group 1). The primary sources of IL-2 were CD4 T cells and B cells. While the contribution of CD4 T cells to IL-2 production was slightly increased with prior exposure though this did not reach statistical significance. IL-6, TNFα, IL-1β and IL-8 were produced largely by CD56^- innate cells (presumably monocytes and dendritic cells), consistent with the canonical innate functions of these cytokines, with no difference in this pattern regardless of pre-exposure. CD4 T cells had the greatest contribution to IL-10 production in unexposed individuals (Group 1), but CD56^- innate cells became the predominant source in individuals with prior exposure (higher proportion in Group 2 + 3 + 5 compared to Group 1, p = 0.005).

In addition to cytokine production, IGRA-positive individuals also showed enhanced levels of activation markers expression in response to C. burnetii antigen stimulation of PBMCs, which was significant for CD69 and CD137 ( Figure 5A ). A more granular analysis of the different exposure groups showed that relative to the control Group 1, C. burnetii-induced CD69 and CD154 expression were significantly elevated in seropositive Group 3 individuals only, while the proportion of cells expressing the activation marker CD137 was significantly increased in both seropositive Groups 3 and 5 ( Figure 5B ).

Finally, we addressed the question of which specific cellular phenotypes characterize the innate and adaptive recall responses induced by C. burnetii antigen stimulation of PBMCs from pre-exposed individuals. To this end, FlowSOM clustering (34) identified 20 sub-populations (meta-clusters C1-C20) within each of the four major populations (CD4 T cells, CD8 T cells, B cells and Innate cells) amongst CD45⁺CD66b^- cells ( Figures 6A, C , 7A, C ). We then assessed whether meta-cluster abundance was increased upon stimulation with C. burnetii antigen (indicating C. burnetii-specific responses, Figures 6B, D , 7B, C ) and whether the specific responses (assessed as background-corrected abundances) were higher in IGRA-positive (Group 2 + 3 + 5) individuals compared to Group 1 controls, and thus linked to prior exposure ( Figure 8 ).

Several meta-clusters amongst both CD4 and CD8 T cells showed enhanced abundance upon stimulation with C. burnetii antigen. CD4 and CD8 meta-clusters exhibiting C. burnetii-specific production of the innate cytokines IL-1β and IL-8 alone (CD4 C2, CD8 C2 and CD8 C3) or in combination with IL-6 and TNFα (CD4 C1 and CD8 C1) did not increase with prior exposure status, or were only found in unexposed individuals (CD4 C3 IL-8⁺) ( Figure 6 ). In contrast, the relative abundances of two CD4 T cell meta-clusters producing adaptive cytokines were significantly higher in pre-exposed individuals: CD4 C11 (CD137⁺CD154⁺CD69⁺IFNγ⁺IL-2⁺TNFα⁺) and C12 (CD137⁺CD154⁺CD69⁺) ( Figure 8 ), consistent with the elevated production of these cytokines by CD4 T cells in Group 5 individuals, as determined by flow cytometry ( Figure 2 ). Amongst CD8 T cells, only C11 (CD137⁺) showed a higher relative abundance in pre-exposed individuals. However, this CD8 meta-cluster showed no effector cytokine production, and the difference in abundance between C. burnetii antigen-stimulated versus unstimulated samples was very minimal, especially in view of the abundance of this meta-cluster in unstimulated PBMCs ( Figure 6B ).

In B cells, eight of the 20 meta-clusters (C3, C19, C16, C20, C6, C12, C9 and C15) showed an increased abundance upon stimulation with C. burnetii antigen in all individuals. The majority of these C. burnetii antigen-stimulation associated B cell meta-clusters was again characterized by production of the innate cytokines IL-1β and IL-8 alone (C19, C6, C12, C9, C15) or in combination with TNFα (C20) or TNFα and IL-6 (C16), and for all relative abundance was unrelated to prior exposure. The same was true for the rare B cell meta-cluster C3 with an activated phenotype but lacking cytokine production ( Figures 7A, B ). No B cell cluster showed evidence for pre-exposure related C. burnetii-specific responses.

Amongst innate cells, there was again an increased abundance upon stimulation with C. burnetii antigen in all individuals, regardless of prior exposure, of meta-clusters showing production of IL-8 only (C12 and C13; likely dendritic cells that lack CD14 and CD56 expression and express high levels of HLA-DR and CD11c) or IL-1β and IL-8 (C3, C5, C2 and C4; CD14⁺ monocytes) ( Figures 7C, D ). Three populations of innate cells, however, showed a significantly higher relative abundance in pre-exposed individuals: the innate meta-clusters C7 (CD11c^highCD14⁺HLA-DR⁺IL-1β⁺IL-6⁺IL-8⁺) and C11 (CD11c^intCD14⁺HLA-DR⁺CD69⁺CD137^intCD154^intIL-1β⁺IL-6^lowIL-8⁺), both likely representing monocytes, and a meta-cluster lacking expression of any canonical markers, C19 (CD69⁺CD137^intCD154⁺IFNγ⁺IL-2⁺TNFα⁺) ( Figure 8 ).

When we further focused on the separate subgroups of IGRA-positive individuals, CD4 and CD8 T cell clusters C11 and innate cell cluster C7 all were significantly increased in Group 5 (seropositive, convalescent from symptomatic infection), while CD4 T cell cluster C12 and innate cell cluster C19 were highest in Group 3 individuals (seropositive, asymptomatic) ( Figure 8 ). Background-corrected frequencies for all T cell and innate cell meta-clusters correlated with IFNγ levels released following stimulation of whole blood, and CD4 T cell cluster C11 and C12 frequencies additionally correlated with IL-2 levels ( Supplementary Table 1 ).