Introduction

The vascular endothelium is lined by a gel-like matrix of highly sulphated glycosaminoglycans (sGAGs) attached to core proteoglycans, the so-called endothelial glycocalyx (eGC). This fragile structure shields the endothelium from pathogenic insults and plays a key role in maintaining microcirculatory homeostasis. Specifically, the eGC acts as a negatively charged “firewall” to reduce leukocyte-endothelial interactions [1,2,3,4]. Its carbohydrate-rich matrix provides resistance to water permeability and contributes to the proportion of albumin molecules “reflected” back into plasma by the vessel wall [5, 6]. In addition, the glycocalyx contributes to the regulation of the redox state and is critically involved in mediating shear-induced nitric oxide release and physiologic anticoagulation [2, 7, 8]. Inflammation-induced eGC dysfunction leads to vascular hyperpermeability resulting in oedema formation and organ damage in critically ill and particularly septic patients [9, 10].

The ultimate shared pathway of eGC damage, particularly in bacterial sepsis and COVID-19, seems to involve the activation and release of heparanase (HPSE), a heparan sulphate (HS)-degrading enzyme that is unique to mammals and breaks down HS chains from HS proteoglycans found in the glycocalyx [9,10,11]. However, upstream of HPSE upregulation and release, only a few specific signalling pathways have been identified, all of which are pathophysiologically relevant in sepsis. These include the angiopoietin-1/Tie2 ligand receptor system, tumor necrosis factor (TNF)-alpha/TNF receptor signalling and toll-like receptor (TLR)-2 and − 4 signalling [2, 3, 12,31, 32].

Incubation of ECs with low to supra-physiological concentrations (0.1–10 ng/ml) of IL-6 alone resulted in a weak and dose-independent increase in sGAG content in the endothelial cell supernatant (data not shown), suggesting that damage to the eGC by classical signalling mechanisms is negligible. However, the addition of a physiologic concentration of sIL-6R (50 ng/ml) to simulate trans-signalling showed a doubling of the amount of sGAG in the supernatant in response to an intermediate concentration of IL-6, consistent with a more potent effect of trans-signalling (Fig. 2A). Accordingly, 2D images and 3D reconstructions of immunofluorescence-stained heparan sulphate, the major GAG of the eGC, showed that the intensity and coverage of the EC surface is reduced after incubation with IL-6/sIL-6R (Fig. 2B, C). As systemic trans-signalling dominates in acute inflammatory responses such as sepsis and COVID-19 [33], all further in vitro experiments were performed with the combination of IL-6 (1 ng/ml) and sIL-6R (50 ng/ml).

We then used the much more accurate nano-indentation AFM method to investigate clinically available inhibitors of IL-6 signalling in vitro. Co-incubation with the humanized monoclonal antibody tocilizumab (100 µg/ml), an antibody against the IL-6R that prevents IL-6 from binding to the IL-6R, completely prevented IL-6-induced glycocalyx damage on ECs (Fig. 2D). Similar results were obtained by blocking Janus kinases (JAKs), which act downstream of gp130, with tofacitinib (10 µM, pre-incubation for 24 h) (Fig. 2E).

We have previously shown that in sepsis and COVID-19, the final common pathway of eGC damage appears to be the activation and release of the heparan sulphate-degrading enzyme heparanase [11, 12]. The addition of heparin - a potent heparanase inhibitor - also protected eGC from IL-6-induced damage, suggesting that IL-6 signalling ultimately acts to regulate heparanase (Fig. 2F).

Fig. 2
figure 2

IL-6 trans-signalling causes eGC damage in vitro. (A) The amount of sulphated glycosaminoglycans (sGAGs) in the supernatant of endothelial cells was measured using the 1,9-dimethylmethylene blue (DMMB) assay. Cells were incubated for 60 min with the indicated concentrations of IL-6 ± sIL-6R, TNFα or LPS supplemented with CD14 (10 ng/ml) and LBP (100 ng/ml). Data are expressed as mean ± SEM of duplicates, n = 3–9. (B + C) Immunofluorescence imaging showing the distribution and coverage of heparan sulphate (green) staining on the cell layer in 3D reconstruction (B) derived from 2D images (C). Nuclei were stained with DAPI (blue). Scale bar 10 μm. Incubation was performed under the same conditions as described in (A) with IL-6 (1 ng/ml) + sIL-6R (50 ng/ml). Incubation in Hepes buffer supplemented with 1% fetal calf serum was used as a control. (D-F) Nanoindentation experiments with atomic force microscopy (AFM) showing changes of eGC height on living endothelial cells after 60 min incubation with sIL-6R + IL-6 ± tocilizumab, tofacitinib or heparin. Incubation in Hepes buffer supplemented with 1% fetal calf serum was used as control. All conditions were pre-incubated for 24 h with vehicle or in case of tofacitinib treatment with tofacitinib. Data are presented as mean ± SEM, each point represents the average of ≥ 4 indentations per cell with a minimum of 8 cells per experiment, n = 3-4. Significance was tested by nested ANOVA followed by Tukey’s or Dunnett’s post-hoc test. *=p < 0.05, **=p < 0.01

Full size image

Tocilizumab protects from serum-induced eGC damage in bacterial sepsis and COVID-19

To simulate more realistic inflammatory conditions in the context of sepsis and COVID-19, three randomly selected serum samples from the third tertile (highest IL-6 concentrations) of each COVID-19 and sepsis were pooled and used for the following in vitro experiments. Mean [± SEM] IL-6 concentrations in the pooled sepsis subgroup were higher than in the COVID-19 subgroup (2859 [± 1275] vs. 636 [± 476] ng/ml). Incubation of ECs with pooled serum (5% for 60 min) resulted in a significant decrease in eGC in both sepsis and COVID-19. Co-incubation with tocilizumab (100 µg/ml) protected the eGC from COVID-19 serum-induced damage (Fig. 3A). A similar trend was observed with co-incubation of tocilizumab with sepsis serum, although it was not statistically significant (Fig. 3B).

Fig. 3
figure 3

Tocilizumab protects from serum-induced eGC damage in bacterial sepsis and COVID-19. Nanoindentation experiments with atomic force microscopy (AFM) showing changes of eGC height on living endothelial cells after 60 min incubation with pooled serum samples (5%) from (A) COVID-19 and (B) sepsis patients with or without concomitant addition of tocilizumab. Pooled sera from healthy individuals were used as control. Data are presented as mean ± SEM, each point represents the average of ≥ 4 indentations per cell with a minimum of 8 cells per experiment, n = 3–4. Significance was tested by nested ANOVA followed by Tukey’s post-hoc test, *=p < 0.05, **=p < 0.01

Full size image

Proteome-derived eGCsignature correlates with IL-6 and outcome in external validation set

In the last step, we wanted to validate the mechanistic link between PBR and IL-6. As our study was not designed for outcome analysis, we used the external MGH COVID-19 cohort, which annotates proteome and outcome data. As sublingual microscopy was not performed in the MGH cohort, we used a previously validated proteomic signature that correlates well with eGC thickness as a surrogate (hereafter referred to as eGCsignature) [15]. Patients’ eGCsignature values correlated inversely well with IL-6 levels (rs = -0.58, p < 0.0001) (Fig. 4A). When dichotomized by the median, patients with lower IL-6 levels had a higher eGCsignature, values indicating a healthier eGC (p < 0.0001) (Fig. 4B). For the composite endpoint of 28-day mortality and/or intubation, eGCsignature performed similarly to IL-6 (AUC [95% CI] 0.86 [0.81–0.91], p < 0.001] vs. IL-6 0.88 [0.83–0.92], p < 0.001), further supporting the proposed mechanistic role of IL-6 in eGC damage (Fig. 4C).

Fig. 4
figure 4

Proteome-derived eGCsignature correlates with IL-6 and outcome in external validation set. External validation of a previously identified proteomic signature that correlates well with eGC thickness (hereafter referred to as eGCsignature) in an independent COVID-19 cohort from Massachusetts General Hospital (n = 219). (A) Dot plot analysis showing the correlation between the eGCsignature and IL-6. (B) eGCsignature classified based on median IL-6 levels. (C) Receiver operating characteristic curves showing the predictive ability of eGCsignature and IL-6 to predict the composite endpoint of 28-day mortality and/or intubation. AU arbitrary units. **p < 0.01

Full size image

Discussion

This study demonstrates that IL-6 and its downstream signalling have a causal role in eGC damage. In vitro pharmacological IL-6 blockade protected against eGC damage induced by sera from bacterial sepsis and COVID-19 patients. Additionally, IL-6 levels correlated with features of eGC impairment and predicted outcomes in an external COVID-19 cohort. These data clearly suggest that IL-6 may be a significant driver of eGC damage during systemic inflammation.

A first hint suggesting a possible link between IL-6 and the eGC came from Ikonomidis et al. who observed a decrease of the PBR (i.e. improvement of the eGC) upon tocilizumab administration in patients with rheumatoid arthritis [34]. In the case of COVID-19, evidence of eGC damage has been found in numerous studies [35,36,37]. However, the pathophysiological pathways that trigger this damage are not yet fully understood.

By establishing a causal role for IL-6 in eGC damage, we add an important piece of the puzzle to the existing literature. Therapeutic strategies targeting IL-6 have been successful in the treatment of severe COVID-19 disease, making this finding even more exciting, and regulatory authorities have recently approved its therapeutic use [38,39,40,41,42]. A recent meta-analysis [43] showed that IL-6 blockade with tocilizumab works best for moderate to severe COVID-19. Contrary to expectations, secondary infections were not increased with the administration of IL-6 receptor antagonists in COVID-19 patients. However, IL-6R blockade increases the risk of bacterial, viral, and opportunistic infections in rheumatoid arthritis [44] and should therefore not be used in bacterial sepsis.

In our cell culture model, blocking not only binding of IL-6 to IL-6R, but also its downstream JAK/STAT pathway with tofacitinib was sufficient to counteract eGC damage. Clinical studies on hospitalized COVID-19 patients have also demonstrated a survival benefit for this class of substances [45,46,47]. However, as none of the clinical trials on anti-IL-6 therapy in COVID-19 used sublingual microscopy to estimate the eGC properties, we are currently unable to estimate the effect of anti-IL-6 treatment on eGC dimensions in vivo. However, due to the widespread systemic vascular involvement in COVID-19, it appears plausible that anti-IL-6 therapy would have reduced damage to the eGC. This hypothesis is further supported by the strong correlation between IL-6 and eGC signature in the validation cohort.

Although it is currently unclear whether IL-6 inhibition has similar benefits in sepsis, a recent Mendelian randomization analysis suggests that IL-6 receptor blockade is associated with lower mortality in 11,643 sepsis patients of the UK Biobank cohort [48]. Barkhausen et al. reported that in a murine polymicrobial sepsis model, pretreatment with a selective inhibitor of IL-6 trans-signalling increased survival from 45 to 100% in a dose-dependent manner [49]. Similarly, specific inhibition of IL-6 trans-signalling completely prevented death in mice with endotoxic shock [50]. Kang et al. also detected improved survival in a murine endotoxic shock model by counteracting vascular injury upon treatment with anti-IL-6R antibody. They could further demonstrate in a murine burn injury model, that short-term IL-6R inhibition preserved eGC integrity in capillaries visualized by electron microscopy [51]. These studies support the results of our in vitro experiments and emphasize the central importance of IL-6 in bacterial sepsis, which is also reflected in its prominent role as an early diagnostic and prognostic biomarker [52,53,54].

However, it is important to note that the preventative effect of tocilizumab on serum-induced eGC damage was somewhat weaker in bacterial sepsis compared to COVID-19. It is unlikely that this is only due to a pure dose effect of IL-6 (which is about four times higher in sepsis), as tocilizumab was administered in a saturating dose. It is more likely that eGC in sepsis is affected by additional and/or more complex mechanisms, which may explain the lack of clinical efficacy of several targets that (at least in theory) should also diminish heparanase release [55]. Further research is required to determine the role of other harmful mediators in the complex cytokine milieu during systemic inflammation. The aim should be to identify the smallest possible set of key mediators or effectors whose blockade can prevent eGC damage in a non-redundant manner. In the case of COVID-19, the singular blockade of IL-6 already appears to be close to achieving this goal.

As this study is primarily hypothesis-generating, it is important to note some limitations. Firstly, the sample size of this cross-sectional study was rather small and was not suitable for the analysis of clinical outcomes. However, the comprehensive dataset combines serum proteomics with intravital microscopy. Secondly, it was not our intention to make a direct comparison between the two entities, but rather to analyze IL-6 in the context of systemic inflammation, using sepsis and COVID-19 as prototypical diseases. Although the entities were initially pooled, the supergroup analyses clearly show that IL-6 appears to play an important role in both. Thirdly, although routine microbiological sampling was performed in all patients, we cannot exclude the possibility of bacterial superinfections in the COVID-19 group, which may have partially influenced IL-6 levels. However, the prevalence of bacterial co-infections in COVID-19 is considered rather low and our COVID-19 cohort had a low median PCT value (0.6 ng/ml), which argues against overt co-infections. Fourthly, measuring the eGC both in vivo and in vitro presents a challenge due to the fragility of this delicate layer. Therefore, we complemented intravital microscopy and atomic force microscopy with additional detection methods, such as immunofluorescence and ELISA measurements. These methods broadly confirmed the microscopy data. However, it is crucial to acknowledge that determining PBR through intravital microscopy and measuring eGC thickness using AFM have limitations, which must be considered when interpreting the results. The AFM method may only detect the denser parts of the eGC. However, a strong correlation between the two methods has been observed multiple times using matched data [12, 19]. A detailed description of the advantages and disadvantages of these and other methods was published recently [4]. Finally, it is possible that randomly selecting and pooling of 3 sera from the upper percentile of each group for in vitro experiments may have introduced bias. However, this approach has been a valid compromise in our previous AFM studies [12, 19].

Conclusion

Our data reveal a novel mechanistic pathway elucidating endothelial glycocalyx damage in inflammatory states such as bacterial sepsis and COVID-19. Further in vivo studies should validate our findings and demonstrate the protective efficacy of potential therapeutic interventions targeting IL-6 signalling on the endothelial glycocalyx. These trials will not only improve our pathomechanistic understanding but also pave the way for targeted interventions.