Multiplex immunohistochemistry differences between Q fever and atherosclerotic abdominal aortic aneurysms indicate immune suppression

Background Chronic Q fever is a zoonosis caused by the bacterium Coxiella burnetii which can manifest as infection of an abdominal aortic aneurysm (AAA). Antibiotic therapy often fails, resulting in severe morbidity and high mortality. Whereas previous studies have focused on inflammatory processes in blood, the aim of this study was to investigate local inflammation in aortic tissue. Methods Multiplex immunohistochemistry was used to investigate local inflammation in Q fever AAAs compared to atherosclerotic AAAs in aorta tissue specimen. Two six-plex panels were used to study both the innate and adaptive immune system. Results Q fever AAAs and atherosclerotic AAAs contained similar numbers of CD68+ macrophages and CD3+ T cells. However, in Q fever AAAs the number of CD68+CD206+ M2 macrophages was increased, while expression of GM-CSF was decreased compared to atherosclerotic AAAs. Furthermore, Q fever AAAs showed an increase in both the number of CD8+ cytotoxic T cells and CD3+FoxP3+ regulatory T cells. Lastly, Q fever AAAs did not contain any well-defined granulomas. Conclusions These findings demonstrate that despite the presence of pro-inflammatory effector cells, there is an immune suppressive micro environment in Q fever AAA resulting in persistent local infection with C. burnetii.


Introduction 1
Q fever is a zoonosis caused by the Gram-negative intracellular bacterium Coxiella burnetii (C. 2 burnetii), with natural reservoirs in a wide range of wild and domestic animals. In infected animals 3 (such as goats) milk, placenta and birth fluids, can contain this microorganism, which may cause 4 human infections via inhalation. Acute Q fever can present as pneumonia, hepatitis, and isolated 5 fever, yet 60% of cases are asymptomatic. (1) Progression to chronic Q fever occurs in approximately 6 5% of infected individuals (2); people at risk are older, suffer from renal insufficiency or aneurysm, or 7 previously underwent valvular or vascular prosthesis surgery. (3) Chronic Q fever manifests as 8 endocarditis or vascular Q fever, i.e., infection of an abdominal aortic aneurysm (AAA) or vascular 9 prosthesis. (4-6) 10 11 Vascular Q fever has severe clinical consequences. In a population of proven and probable vascular Q 12 fever patients according to the Dutch consensus guideline, a Dutch cohort study has described that 13 complications had occurred in 61% of the cases. Of these complications, acute aneurysms (i.e. 14 rupture, dissection, endoleak or symptomatic aneurysms) were most prevalent (35%), followed by 15 abscesses (22%), and fistula (14%). Moreover, 25% of patients had a definitely or probably chronic Q 16 fever related cause of death. (6) In addition, serological screening of 770 patients with aorto-iliac 17 disease, e.g., aneurysms or previous vascular reconstructions, demonstrated that 16.9% was 18 seropositive for Q fever, of which 30.8% suffered from chronic Q fever. In this group, aneurysm-19 related acute complications were more common than in aneurysm patients without Q fever. (7)  20   21 To elucidate the pathology underlying chronic Q fever, previous studies have mainly focussed on 22 immune responses in peripheral blood. Blood mononuclear cells of patients with chronic Q fever, 23 when exposed to C. burnetii in vitro, produce high amounts of Interferon-gamma (IFNg), the 24 1 Q fever aortas reveal extensive fibrosis 2 All above mentioned features are signs of chronic inflammation and long-existing disease. This was 3 supported by HE-and Elastin Van Gieson (EVG) stainings, which demonstrated destruction of elastin 4 fibers and fibrosis. Both atherosclerotic AAAs and Q fever AAAs exhibited extensive atherosclerotic 5 plaque formation. However, there were large differences in vessel architecture as demonstrated in 6 Figure 6. Earlier studies have extensively described that aortic aneurysms show fragmentation of 7 elastin fibers indicating media degeneration. (15,16) In our series, the amount of elastin fibers was 8 even more decreased in many Q fever AAAs than in atherosclerotic AAAs with a similar diameter 9 (marked with black arrows). In addition, the tunica adventitia showed extensive fibrosis in Q fever 10 AAAs (marked with asterisks in Figure 6I and 6L). These changes indicate the (more pronounced) 11 disrupted architecture in Q fever AAAs, which can attribute to ongoing inflammation. 12

Discussion 1
We are the first to introduce mIHC in vascular Q fever to study ongoing local inflammation. This 2 sophisticated mIHC method enabled us to quantify immune cells in large sections of tissue which 3 minimized sampling bias. First, we showed that granulomas are absent in Q fever AAAs. Second, 4 atherosclerotic and Q fever AAAs were similar when comparing numbers of immune cells. However, 5 there were striking differences in the composition of macrophage-and T cell-phenotypes between 6 AAAs and Q fever AAAs, leading to new insights into the pathogenesis of vascular Q fever and its 7 complications and possibly with therapeutic consequences. 8 9 Our first observation, the absence of well-formed granuloma formation in our cohort of Q fever AAAs 10 is an important one, since it suggests that the local immune landscape lacks an adequate pro-11 inflammatory response. In our series, we could not find any well-formed granuloma similar to how 12 they are described in acute Q fever. In acute Q fever, so called doughnut granulomas are reported: 13 granulomas with a central clear space and a fibrin ring within or at its periphery (1,17), for example 14 in liver biopsies in case of hepatitis. (18) Here, granuloma is a feature of active defense against the 15 pathogen. This contrasts to chronic Q fever, where granulomas have not been described before. (17,16 19) In particular, Lepidi described that resected valve specimens of patients with Q fever endocarditis 17 lacked well-formed granulomas. (12) In vascular Q fever, granulomatous responses consisting of 18 histiocytes surrounding necrotic areas have been reported in Q fever AAAs, however well-formed 19 granulomas were not found. (20) Thus, we would interpret the absence of organized granulomas as 20 the first clue for an immune-suppressed environment in AAA of Q fever patients that allows 21 persisting infection after the acute phase. 22 Secondly, our results demonstrate some similarities between Q fever AAAs and AAAs. Percentages of 1 CD3 + T cells, CD20 + B cells, CD1c + cDC2, CD15 + neutrophils, and CD68 + macrophages are similar 2 between the groups with atherosclerotic AAA and Q fever AAA. This finding is supported by the PCA, 3 which shows overlapping populations of atherosclerotic and Q fever AAAs when entering these 4 inflammatory cell markers. This does not come as a surprise since it is well established that vascular 5 Q fever develops in preexisting atherosclerotic aneurysms. (5,7,(20)(21)(22) 6   7 Despite the similarities, we discovered that atherosclerotic and Q fever AAAs do have important 8 differences, which emerge when investigating macrophage and T cell subset markers. Macrophages 9 in Q fever AAAs were found to be polarized into the less inflammatory M2 phenotype, which is 10 'tolerogenic' and poorly microbicidal, in contrast to the M1 phenotype that possesses a machinery 11 that can clear an infection. Interestingly, in the AAAs we found less M2 polarization. This would 12 either indicate that macrophages polarize towards M2 in response to C. burnetii infection, or that the 13 presence of M2 polarization is a prerequisite for C. burnetii persistence. Previous studies have 14 demonstrated that C. burnetii inhabits and proliferates in monocytes and macrophages, and more 15 specifically, in resident vascular wall macrophages in case of vascular Q fever. bone-marrow derived macrophages after infection with C. burnettii compared to C. burnetii-infected 22 wild type mice. These previous findings suggested that chronic Q fever is associated with M2 23 polarization of macrophages, but direct evidence in chronic Q fever patients was lacking. Our findings 24 establish that outgrowth and persistence of C. burnetii in AAAs is associated with the predominance 1 of M2 macrophages. 2 3 There are several possible explanations for the lack of macrophage activation. First, our results 4 demonstrate decreased expression of GM-CSF in Q fever AAAs compared to AAAs. GM-CSF is a pro-5 inflammatory cytokine that activates granulocytes and macrophages. (24) Its decreased expression in 6 Q fever AAAs may contribute to the immune suppressive environment in Q fever AAA. The role of 7 GM-CSF in the context of aneurysm formation has been investigated previously. (25) Strikingly, Son 8 et al described increased occurrence of aortic dissection and intramural hematoma in wild type mice 9 subjected to aortic inflammation (CaCl2 + Ang II administration) when also receiving GM-CSF. Only 10 administrating GM-CSF, without the prerequisite of aortic inflammation, did not result in aortic 11 dissection or intramural hematoma. Its potential clinical relevance was confirmed in human blood: 12 GM-CSF serum levels of patients suffering from acute dissection were higher than controls with 13 coronary artery disease, aortic aneurysms of healthy volunteers. (25) Additionally, in our cohort we 14 found that Q fever AAAs rupture at smaller diameter compared to atherosclerotic AAA. This finding, 15 combined with the GM-CSF paradox, suggests that the development of Q fever AAAs and 16 atherosclerotic AAAs follow different pathways, however strictly hypothetically. 17 18 Second, a key cytokine in activation is IFNg, a T-helper (Th)-1 cytokine that activates macrophages 19 and makes them more microbicidal. Previous studies from our group have demonstrated that 20 peripheral blood mononuclear cells from patients with chronic Q fever exhibit an abundant 21 production of IFNg when exposed to C. burnetii antigens. (8, 9) These findings were enigmatic since 22 there is an apparent inability of the patient's immune system to kill C. burnetii at the infected sites. 23 The current findings would be compatible with a downregulated IFNg response at the infected site. 24 1 In addition to differences in macrophage subsets, differences in T cell subsets were also observed. 2 First, the number of cytotoxic T cells was increased in both infiltrate and surrounding tissue of Q 3 fever AAA compared to AAA. Although the numbers of cytotoxic T cells were high, their function 4 might be compromised, resulting in defective elimination of C. burnetii. The increased numbers of 5 regulatory T cells we found in Q fever AAAs may play a role here. An increased number of circulating 6 regulatory T-cells has also been shown by Layez et al in Q fever endocarditis patients and in acute Q 7 fever patients. (13) Regulatory T cells can inhibit cytotoxic T cells directly or indirectly (26), with a 8 possible role for IL-10 produced by this T cell subset. An important role of IL-10 in chronic 9 development of Q fever has been postulated based on converging evidence from a series of in vitro 10 studies. IL-10 production by peripheral blood mononuclear cells from patients with Q fever 11 endocarditis and Q fever with valvulopathy who were at risk for developing chronic Q fever was high, 12 compared to control individuals. (27, 28) Moreover, IL-10 specifically increases C. burnetii replication 13 in naive monocytes (29) possibly by downregulating IFNg. Finally, low IL-10 production in monocytes 14 from patients with acute Q fever was associated with C. burnetii elimination, whereas C. burnetii 15 replicated in monocytes from patients with chronic Q fever and high IL-10 production. The 16 microbicidal activity of monocytes from patients with chronic Q fever was restored by neutralizing IL-17 10. (30) The murine model of chronic Q fever mentioned above, also confirmed a key role for IL-10 in 18 bacterial persistence. C. burnetii infection is persistent in mice that overexpress IL-10 in the 19 macrophage compartment. (23) Thus, IL-10 could play a crucial role in this immune-suppressed 20 environment. 21

22
The last major difference between Q fever AAA and atherosclerotic AAA is the extent of damage to 23 the vascular wall architecture in Q fever AAAs. This is demonstrated by extensive loss of elastin fibers 24 and increase of fibrosis present in the vascular wall. Fragmentation of elastin fibers has been 25 described for AAAs previously (16), however we found the loss of elastin fibers more evident in the 1 lesions from Q fever AAAs than atherosclerotic AAAs. Fibrosis is characterized by replacement of 2 normal tissue by excessive connective tissue, and usually follows chronic inflammation. This may be 3 the effect of persistent presence of growth factors, proteolytic enzymes, angiogenic factors, and 4 profibrotic cytokines. (31, 32) Previously, fibrosis was also observed in chronic Q fever endocarditis in 5 humans and cows. (12,(33)(34)(35) This indicates that our Q fever AAA cohort suffered from more 6 destructive disease than our AAA cohort. 7 8 These novel insights could lead to new clues for novel treatments and thus developments for clinical 9 care. Currently, Q fever AAA still leads to significant morbidity and mortality rates despite antibiotic 10 and surgical treatment. Epidemiological studies demonstrate the similar risk profile of Q fever and 11 non-Q fever infected AAAs, yet the risk of complications is higher in the Q fever infected group (7), 12 even up to 61% (6), and 25% of patients suffering from Q fever AAA had deceased with a 13 definitely/probably chronic Q fever related cause of death. (6) Here, we confirm that in vascular Q 14 fever, the local immune response is skewed towards an immunotolerant state. Hypothetically, the 15 decreased expression of GM-CSF suggests a possible role for immunomodulating treatment, for 16 example with administration of recombinant GM-CSF. This is already approved for neutropenia due 17 to myelosuppression (36), and has been suggested for treatment for pulmonary tuberculosis. (37)  18 There might be a role for immunomodulating adjuvant therapies in patients with Q fever AAA in 19 whom treatment failure is observed with antibiotics alone. 20

21
Our study was the first to use mIHC in Q fever AAA and thereby to gain information about the 22 number and proportion of immune cells, and simultaneously obtain spatial information. This 23 powerful technique and the access to rare Q fever AAA tissue are strengths of this study. While other 24 studies have tested for immune cell activation and recruitment in peripheral blood, we were able to 25 study the actual infected tissue. Interpreting our results in context of previous observations enables 1 us to increase our understanding of the pathophysiology of Q fever AAA. Still, several limitations 2 should be noted. First, our sample size is limited with only ten vascular Q fever samples. However, 3 this is still the largest study investigating local immune responses in Q fever AAA in humans. In 4 addition, in our quantification method we include entire slides up to 238 20x views per patient, 5 which minimizes the effects of the small sample size. Second, consistent with IHC studies in general, 6 we can only describe the immune cells we observe, without answering mechanistic questions. 7 Nevertheless, when interpreting our results in light of the current literature, we can reasonably 8 formulate hypotheses about the pathophysiology and test these in further research. 9 10 Taken together, this leads to the following model with a prominent role for immune suppression. 11 First, macrophages that harbor C. burnetii are not effectively killing the organisms, probably due to a 12 lack of activation by proinflammatory cytokines like GM-CSF and IFN-g in a microenvironment with 13 excess IL-10. Second, effector T cells that attempt to eliminate the intracellular bacterium residing in 14 monocytes and macrophages, are hindered by regulatory T cells that are prominent IL-10 producers. 15 Third, there is a lack of microbicidal M1 macrophages, instead macrophages are polarized into the 16 tolerogenic M2 phenotype, which leads to insufficient attack of the pathogen, enabling persistent 17 infection. 18

Methods and materials 1
Abdominal aorta tissue samples from patients with Q fever infected aneurysms and control groups 2 were investigated with a novel mIHC method to study the involvement of the innate and adaptive 3 immune system in vascular Q fever. The data underlying this article will be shared upon reasonable 4 request to the corresponding author. During surgery, the ventral part of the abdominal aorta was removed. If necessary, adhering thrombi 3 were gently removed from the tissue before further processing. Directly after collection, samples 4 were fixed in buffered 4% formaldehyde for at least 24 hours and no longer than 72 hours. If large 5 amounts of calcification were present, samples were decalcified by storing them in EDTA solution for 6 another 24 hours. Subsequently, samples were carefully embedded in paraffin in an attempt to 7 include all aorta layers (formalin fixed and paraffin embedded (FFPE)). Of these tissues, full thickness 8 transverse sections of 4 µm were mounted on silane coated glass slides (New Silane III, MUTO PURE 9 CHEMICALS, Japan). 10 11 Multiplex Immunohistochemistry 12 Samples were stained with two mIHC panels, which enclosed the innate and adaptive immune 13 system (Table 2). Optimization and validation of mIHC panels were performed as described 14 previously. (40) Samples were stained with six consecutive tyramide signal amplification (TSA) stains 15 followed by antigen stripping after every staining. This resulted in the fluorophore remaining on the 16 target, thus enabling eight simultaneous colors on one slide (six markers, DAPI and 17 autofluorescence). Slides were stained automatically in a Leica Bond system (BOND-Rx Fully 18 Automated IHC and ISH, Leica Biosystems). After positioning in the machine, slides were 19 deparaffinized, rehydrated and washed with demi water. After this, samples underwent heat induced 20 antigen retrieval (HIER) in BOND Epitope Retrieval 2 (AR9640, Leica Biosystems) or BOND Epitope 21 Retrieval 1 (AR9961, Leica Biosystems) during 20 minutes for the adaptive and innate panel, 22 respectively. Then, protein blocking in Akoya Antibody Diluent/Block (Akoya biosciences, MA) took 23 place for 10 minutes, followed by incubation with the first primary antibody for 1 hour, subsequently 24 with the secondary antibody (Polymer HRP, Ms + Rb (Akoya biosciences, MA)) for 30 minutes and 25 finally with an Opal fluorophore ((Akoya biosciences, MA) dissolved 1:50 in 1 X Plus Amplification 1 Diluent (Akoya biosciences, MA) for 10 minutes. To facilitate multiplex staining with six markers, 2 samples were heated for 10 minutes which enabled antigen stripping. After this staining cycle, this 3 procedure was repeated for five different primary antibodies, the secondary antibody and 4 corresponding Opal fluorophores. Finally, DAPI was used as a nuclear counterstain and slides were 5 mounted with Fluoromount-G (0100-01; Southern Biotech, Birmingham, AL, USA). All incubations 6 steps were performed at room temperature. Please see Supplementary Table 2 Figures 1A and 2A). 17 Subsequently, inForm Advance Imaging Analysis software was used for segmentation of tissue and 18 cells. For tissue segmentation, tissue slides were divided into tissue, infiltrate, thrombus, blood, and 19 background (Supplementary Figure 1C). This segmentation was based on DAPI, autofluorescence 20 and, if present, also CD20 and CD3. For single cell segmentation, cells were identified with DAPI and 21 autofluorescence, and, depending on the panel, with membrane markers CD20 and CD3 22 (Supplementary Figure 1B) or CD68, CD206 and CD15. Requiring DAPI for cell segmentation ensured 23 the exclusion of artifact staining, in which DAPI is absent. The output of the software was 20x 24 magnification images and cell data (localization, tissue, phenotype, and marker) per slide. Images 25 were combined into single flow cytometry standard (fcs) files, allowing analysis in FlowJo (FlowJo 1 10.0.7, Becton Dickinson, NJ). In FlowJo, only cells in tissue and infiltrate were analyzed and gates 2 were drawn as shown in Supplementary Figure 1E for the adaptive panel and Supplementary Figure  3 2B for the innate panel by two observers with excellent interobserver correlation (Supplementary 4 Figure 1D and 2D). 5 These clear distinct positive cell populations were not found for CD45RO and MMP9 as their 6 expression is gradual. (41) For that matter, the gates for these markers were drawn in negative 7 populations, namely non-T cells and non-neutrophils, respectively. Following this, these gates were 8 copied to populations that could express these markers. GM-CSF fcs files did not show distinct 9 positive and negative populations although they were visible at the microscopy images. Therefore, 10 inForm Advance Imaging Analysis software was used for automatically thresholding for GM-CSF per 11 sample, providing the number of GM-CSF positive pixels per sample (Supplementary Figure 2C). Corp) was used for statistical analysis. PRISM 8.0.2 (Graphpad, GSL Biotech LLC, CA) was used for 20 visualization of results. Continuous data were expressed as mean ± standard deviation (SD), or in 21 case of non-Gaussian distribution, as median (interquartile range) (IQR). Kruskal-Wallis test adjusted 22 with Bonferroni correction for multiple testing was used for testing continuous variables between 23 four groups. Binary variables were tested for differences using the Fisher exact test. Interobserver 1 variability was calculated with the intraclass correlation coefficient (ICC). Correlations between 2 continuous non-Gaussian distributed variables were studied with Kendall's tau because of low 3 numbers per group. P < 0.05 was considered statistically significant. Principal component analysis 4 was performed in RStudio 1.2.5033 (RStudio, inc. Boston, MA) and R (R Foundation for Statistical 5 Computing, Vienna, Austria) using singular value decomposition and the tidyverse (42) and factoextra 6 packages. (43)  7 8 Acknowledgements 9 The authors would like to acknowledge Anne van Duffelen and Kiek Verrijp for their help in 10 optimizing the staining procedure, Jelena Meek for assistance in staining, Inge Wortel, Shabaz Sultan, 11 and Johannes Textor for their help in data analysis, and Janneke Timmermans for her feedback.    In the adventitia (I, L) tissue is replaced by large amounts of fibrosis, indicated with asterisks.