Dipeptidylpeptidase (DPP)-4 inhibitor therapy increases circulating levels of anti-inflammatory soluble frizzle receptor protein (sFRP)-5 which is decreased in severe COVID-19 disease

Obesity and type 2 diabetes (T2D) show an increased risk for a severe COVID-19 disease. Treatment with DPP4 inhibitor (DPP4i) results in reduced mortality and better clinical outcome. Here, we aimed to identify potential mechanisms for the observed DPP4i effect in COVID-19. Comparing T2D subjects with and without DPP4i treatment, we identified a significant increase of the anti-inflammatory adipokine sFRP5 in relation to DPP4 inhibition. sFRP5 is a specific antagonist to Wnt5a, a glycopeptide secreted by adipose tissue macrophages acting pro-inflammatory in various diseases. We therefore examined sFRP5 levels in patients hospitalised for severe COVID-19 and found significant lower levels compared to healthy controls. Since sFRP5 might consequently be a molecular link for the beneficial effects of DPP4i in COVID-19, we further aimed to identify the exact source of sFRP5 in adipose tissue on cellular level. We therefore isolated pre-adipocytes, mature adipocytes and macrophages from adipose tissue biopsies and performed western-blotting. Results showed a sFRP5 expression specifically in mature adipocytes of subcutaneous and omental adipose tissue. In summary, our data suggest that DPP4i increase serum levels of anti-inflammatory sFRP5 which might be beneficial in COVID-19, reflecting a state of sFRP5 deficiency.


Results
DPP4i treatment is related to increased serum concentrations of the anti-inflammatory adipokine sFRP5 in human T2D subjects. The aim of the present study was to examine, if DPP4i exert antiinflammatory effects in treated T2D subjects which might explain the beneficial effects in Covid-19 patients. Therefore, differences in clinical and biochemical parameters between diabetic subjects with DPP4i treatment (T2D + DPP4i), without DPP4i treatment (T2D-DPP4i) and healthy controls from our FoCus cohort were investigated. While most of the tested variables were not different between the groups, a significant increase of the anti-inflammatory adipokine sFRP5 in diabetic subjects taking DPP4i compared to diabetic subjects without DPP4i treatment and healthy controls was found (p = 0.013; p = 0.0021, Fig. 1E).This is of interest, since the Wnt5a/sFRP5 system has been shown not only to be important in metabolic inflammation but also in severe bacterial sepsis 7 . In the present analysis, the pro-inflammatory counterpart, Wnt5a did not reveal significant differences between the groups (Fig. 1F). Besides sFRP5, parameters of glucose status: glucose and HOMA-IR (Fig. 1A,B), inflammatory markers: IL-6 and CRP (Fig. 1C,D) showed significant differences between healthy controls and diabetes subjects, but not between subjects with and without DPP4i treatment. Comorbidities, lipid profile, metabolites and anthropometry did not show significant differences between the diabetic subjects with and without DPP4i medication. Descriptive statistics and results of associations with clinical markers are shown in Tables 1 and 2. Serum concentrations of sFRP5 are reduced in severe COVID-19 disease. Having found that anti-inflammatory sFRP5 is increased in DPP4i treated T2D subjects we next examined, if sFRP5 and Wnt5a are altered in COVID-19 patients. These examinations were performed in a cohort of 17 patients hospitalised on intensive care unit for severe COVID-19 and compared to healthy controls from our FoCus cohort. Table 3 displays basic characteristics of the Covid-19 subjects from the ICU-cohort and healthy controls from the FoCus cohort. Of interest, we found that subjects with COVID-19 revealed significant lower sFRP5 levels compared to healthy controls (p < 0.01, Fig. 2A). Moreover, Wnt5a levels were significantly increased in COVID-19 patients compared to healthy controls (p < 0.001, Fig. 2B). sFRP5 is specifically expressed in mature adipocytes. So far, we found that DPP-4 inhibitor treatment is associated with increased sFRP5 serum concentrations and that COVID-19 is related to sFRP5 deficiency. sFRP5 is known to be an adipokine. In a next set of experiments, we aimed to examine, (1) which specific cell type in adipose tissue is important for sFRP5 synthesis and release in order to identify the potential DPP4i cellular target and (2) whether sFRP5 is differentially expressed in subcutaneous (SAT) versus visceral adipose tissue (VAT), since it is known that inflammatory reactions are more closely linked to visceral adipose tissue. To this end we isolated pre-adipocytes, mature adipocytes and macrophages from adipose tissue biopsies from n = 100 human subjects undergoing elective abdominal surgery (mainly bariatric surgery). The basic characteristics of the adipose tissue biopsy cohort are given in Table 4.
Protein expression of sFRP5 was investigated by western blotting. sFRP5 was not detectable in pre-adipocytes or macrophages. In contrast, mature adipocytes revealed a significant sFRP5 expression (Fig. 3) suggesting sFRP5 is an adipokine specifically secreted by fully differentiated adipocytes in humans.
Of interest, although inflammation is more linked to VAT, sFRP5 expression was found in both, mature adipocytes from subcutaneous and visceral adipose tissue biopsies with no significant difference.

Discussion
Patients with T2D and severe COVID-19 disease showed better clinical outcomes with intake of the DPP4i sitagliptin 6 . sDPP4 activity was described to be elevated during infections with hepatitis C virus or Epstein-Barr virus 8 . DPP4 cleaves the pro-inflammatory chemokine CXCL10 leading to the short NH2-terminal truncated CXCL10 form which is positively associated with hepatitis C virus infection [9][10][11] . Treatment with DPP4i reduced cleavage of CXCL10 12 which points out that sDPP4i exhibit more therapeutic functions than increasing GLP-1 levels.
To examine further potential mechanisms of the anti-viral/anti-inflammatory DPP4i activity, we compared T2D patients treated with DPP4i (n = 23), without DPP4i (n = 46) and healthy controls (n = 46) on various clinical and biochemical levels. Significant higher levels of the Wnt5a inhibitor sFRP5 in T2D subjects with DPP4i treatment compared to T2D subjects without DPP4i treatment and healthy controls were found. Metabolic markers such as glucose and HOMA-IR and the inflammatory parameters CRP and IL-6 showed significant higher www.nature.com/scientificreports/ levels in diabetic subjects with and without DPP4i treatment compared to healthy controls. However, diabetic subjects with and without DPP4i medication had no differences in these clinical parameters which points out the independent elevation of SFRP5 by DPP4 inhibition. Our previous studies showed that the pro-inflammatory Wnt5a/anti-inflammatory sFRP5 system is closely related to various acute (e.g. sepsis 7 ) as well as chronic (e.g. psoriasis 13 ) inflammatory diseases and is also important in the context of metabolic inflammation in obesity and type 2 diabetes. Thereby, obese subjects reveal reduced concentrations of anti-inflammatory sFRP5 and increased levels of Wnt5a 14,15 . In our previous clinical study in human subjects suffering from bacterial sepsis we observed stable (not reduced) sFRP5 concentrations 7 . This is of interest, since in the present analysis sFRP5 concentrations were , CRP (C), IL-6 (D) were significantly higher in diabetic subjects with and without DPP4i treatment compared to healthy controls, no differences between subjects with and without DPP4i treatment. (E) Diabetic subjects with DPP4i treatment showed significant higher levels of sFRP5 compared to diabetic subjects without DPP4i treatment and healthy controls. (F) Wnt5a levels showed no differences between groups. Wilcoxon test; statistical significance p < 0.05. Wnt5a is a glycoprotein and displays a regulator of the innate immune response by reinforcing low-grade inflammation. It is secreted by pro-inflammatory macrophages which infiltrate adipose tissue. Previous investigations of our working group showed in a FoCus subcohort of n = 896 cross-sectional subjects a positive association of Wnt5a with the clinical markers IL-6 and triglycerides. Furthermore, subjects with T2D exhibited high levels of Wnt5a which were positively correlated with fasting plasma glucose levels 15 . sFRP5 is a soluble Wnt5a receptor and inhibits pro-inflammatory effects of Wnt5a by sequestering it extracellularly. Accordingly, sFRP5 has been indicated as potential therapeutic target in metabolic inflammation 7,15,16 .

Scientific Reports
Having found DPP4i treatment to be related with increased sFRP5 serum levels, we specifically analysed differences of serum sFRP5 levels between subjects with severe COVID-19 disease and healthy Focus controls. We observed significant lower sFRP5 levels in COVID-19 patients compared to healthy controls, while Wnt5a levels were significantly increased in COVID-19 patients. Of interest, Choi et al. recently also showed significantly increased Wnt5a serum levels in severe COVID-19 infection 17 . Hence, results point towards the activation of wnt signalling pathway in COVID-19, that might be even more pronounced in severe inflammation resulting from SARS-CoV-2 rather than classical bacterial septic infection due to the additional reduction of anti-inflammatory sFRP5 in COVID-19 but not in bacterial sepsis. Furthermore, it might be of interest how sFRP5/wnt5a levels act in subjects with mild COVID-19 to prevent a possible more severe course of disease or post covid-19 symptoms.
To further investigate sFRP5 characteristics in metabolic inflammation, we analysed differential expression of sFRP5 on a cellular level in pre-adipocytes, adipocytes and macrophages. Results showed sFRP5 expression specifically in visceral and subcutaneous mature adipocytes while for pre-adipocytes and macrophages, no sFRP5 expression was recorded. These results confirm the findings of other working groups 18,19 which indicates that compared to Wnt5a, sFRP5 reflects a specific marker of mature adipocytes. Furthermore, Wang et al. showed that sFRP5 expression showed higher levels in VAT from obese patients compared to lean controls 18 . These findings contradict the assumption that sFRP5 is only expressed in healthy adipocytes and reveals decreased levels in obesity 20 . It is also to be mentioned that sFRP5 is not only expressed in adipose tissue but also in pancreatic islets in which sFRP5 and Wnt signalling were shown to be regulators of β-cell proliferation 21 . Since it was observed that a SARS-Cov-2 infection can induce β-cell damage and therefore reinforce the development of T2D 22 , it is of importance to investigate the effects of DPP4i not only in the adipose tissue but also in pancreatic islets.
Thus, the results of our present study together with our previous results in obese subjects 15 might explain obesity being a risk factor for severe disease courses of COVID-19, since sFRP5 levels are lower in both, COVID-19 www.nature.com/scientificreports/ and severe obesity. However, a limitation of this analysis must be mentioned that the BMI of Covid-19 subjects could not be determined because they were in intensive care. Only two subjects were diagnosed as obese. Besides Wnt5a scavenging, the specific mechanisms of sFRP5 are still barely investigated. According to its anti-inflammatory properties, Carstensen-Kirberg et al. observed that sFRP5 decreased levels of Il-6 and NF-κB phosphorylation in TNFα treated human adipocytes 23 , further suggesting anti-inflammatory activities.
In summary, our data suggest that the release of the anti-inflammatory adipokine sFRP5 from mature adipocytes in humans is induced by DPP4i which might compensate the sFRP5 deficiency found in COVID-19, but not in bacterial sepsis. In addition, it might be speculated that elevated levels of sFRP5 might be an indication for a better clinical outcome of COVID-19 in obese as well as T2D patients, which could be interesting to examine in future biomarker studies.

Statistical analysis. Statistical analysis was performed using open source R software and SPSS Statistics
Software Version 24. Statistical significance was set at p < 0.05. From FoCus cohort, a subcohort of patients was divided into subjects with type 2 diabetes taking DPP4i (n = 23; "T2D + DPP4i"), without DPP4i treatment (n = 46; "T2D-DPP4i") and healthy subjects (n = 46; healthy controls). Groups were matched by age and gender. Continuous variables were tested for normal distribution using the Shapiro-Wilk test. To determine between-group differences, Wilcoxon rank sum test was performed. Categorical variables are shown as absolute numbers and percentages. x 2 test and Fisher test were used for between-group analyses.
The ICU cohort contained 25 subjects. Since recalls were excluded, 17 Covid-19 subjects of the ICU cohort were selected and matched by gender and age with healthy controls of FoCus cohort (n = 34). Since Covid-19 subjects were in intensive care, the BMI could not be determined. However, 2 ICU subjects were diagnosed as obese. Healthy controls had a BMI lower than 25. sFRP5 and Wnt5a differences between groups were determined using Wilcoxon rank sum test.
For descriptive statistics, variables of adipose tissue cohort were tested for normal distribution using the Shapiro-Wilk test and are expressed as median (IQR) according to non-normal distribution. Mann-Whitney U test were used to determine group-differences.

Data availability
The datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request.