NETosis‐associated serum biomarkers are reduced in type 1 diabetes in association with neutrophil count

Summary

As the immune pathways involved in the pathogenesis of type 1 diabetes (T1D) are not fully understood, biomarkers implicating novel mechanisms of disease are of great interest and call for independent evaluation. Recently, it was reported that individuals with T1D display dramatic elevations in circulating components of neutrophil extracellular traps (NETs), indicating a potential role for NETosis in T1D. Our aim was to evaluate further the potential of NET‐associated proteins as novel circulating biomarkers in T1D. We tested serum from subjects with T1D (n = 44) with a median age of 26·5 years and a median duration of 2·2 years, along with 38 age‐matched controls. T1D subjects did not show elevations in either neutrophil elastase (NE) or proteinase 3 (PR3), as reported previously. In fact, both NE and PR3 levels were reduced significantly in T1D subjects, particularly in subjects within 3 years of diagnosis, consistent with the known reduction in neutrophil counts in recent‐onset T1D. Indeed, levels of both NE and PR3 correlated with absolute neutrophil counts. Therefore, while not ruling out potential local or transient spikes in NETosis activity, the levels of these serum markers do not support a role for systemically elevated NETosis in the T1D population we studied. Rather, a modest reduction in these markers may reflect other important aspects of disease activity associated with reduced neutrophil numbers.

Introduction

Type 1 diabetes (T1D) is an organ‐specific autoimmune disease characterized by elevated blood glucose resulting from the destruction of insulin‐producing beta cells in the pancreas. The role of adaptive immunity in T1D has been well established by the strong genetic association with certain human leucocyte antigen (HLA) haplotypes and clear loss of tolerance to beta cell antigens in both B and T cells 1, 2. Indeed, HLA genetics and serum autoantibodies have each served as highly informative biomarkers for predicting risk of developing T1D, allowing for the study of these individuals in a prospective fashion, including in preventative intervention trials. Biomarkers that can predict the course of disease after diagnosis, however, have lagged significantly behind those in the pre‐diabetes setting.

There is currently a significant effort to identify immune parameters that correlate with rate of beta cell loss after T1D onset, and much of this effort continues to be focused on adaptive immunity 3. However, there is also increasing interest in the role of innate immunity and inflammation in T1D 4. For instance, the frequency of natural killer (NK) T cells and a subpopulation of myeloid dendritic cells has been reported to correlate with C‐peptide loss 5, 6. In addition, multiple laboratories have reported a reduction in the number of circulating neutrophils near the time of T1D diagnosis as well as prior to onset 7, 8. Somewhat paradoxically considering these findings, a recent study has raised a significant amount of interest in the field, showing extremely high levels of neutrophil‐derived enzymes in the serum of newly diagnosed T1D patients 9. Specifically, these products were associated with NETosis, a neutrophil‐specific form of anti‐microbial cell death.

NETosis is an activation‐induced process in which neutrophils extrude neutrophil extracellular traps (NETs), consisting of de‐condensed chromatin associated with granules of degradative enzymes such as neutrophil elastase (NE) and proteinase 3 (PR3), through the plasma membrane for the purpose of preventing migration of extracellular pathogens and facilitating their destruction 10. In addition to its role in bacterial immunity, there is increasing evidence that NETosis may also contribute to the pathogenesis of autoimmune diseases, including rheumatoid arthritis (RA), systemic lupus erythematosus (SLE) and vasculitis 11, 12. In anti‐neutrophil cytoplasmic antibody (ANCA)‐positive vasculitis and SLE, there is direct evidence for production of autoantibodies targeting NET‐associated proteins 13, while anti‐citrullinated protein autoantibodies (ACPA), which are linked closely to and predictive of RA, may occur in response to the post‐translationally modified protein content in NETs 14. Immune responses against antigens that are post‐translationally modified, including by citrullination, have also been reported in T1D 15, 16. Whether these autoantigen modifications have any dependence on NETosis or are unrelated events occurring in beta cells or antigen‐presenting cells is not known.

In this study, we sought to confirm the recent report of elevated NETosis‐associated factors in the serum of subjects with T1D 9 in order to explore further the utility of this biomarker for predicting clinical outcomes in T1D.

Materials and methods

Human subjects

Blood samples were collected from subjects with T1D enrolled in the Benaroya Research Institute Diabetes Translational Research Project (BRIDge) study and from control subjects without diabetes participating in the Immune‐Mediated Diseases Registry and Repository: Control Group Translational Research Study. T1D cases were defined using the standard ADA criteria. Blood draws and research on stored samples were performed with the written consent of the participants and with the approval of the BRI Institutional Review Board. Forty‐four subjects with T1D with a median age of 26·5 years and median disease duration of 2·2 years and 38 control donors were identified from the BRI sample repository. Groups were matched for age, gender, ethnicity and date of draw (duration in storage). Clinical information, such as HbA1c, complete blood count (CBC) and islet‐specific autoantibody results, were also obtained from the majority of T1D subjects.

Measurement of NETosis enzymes

Venous blood was collected in tiger‐top serum separator tubes (SST) via 19‐ or 21‐gauge butterfly needle, allowed to clot at room temperature for at least 30 min, spun, aliquoted and frozen at −80°C, where they were stored until use. Serum was frozen typically within 2 h of blood draws. Serum levels of PR3 and NE were measured using commercial enzyme‐linked immunosorbent assay (ELISA) kits from BioVendor (Asheville, NC, USA) and eBioscience (San Diego, CA, USA), respectively. Serum samples were thawed at room temperature, diluted 100‐fold in the supplied assay buffer and then added to precoated ELISA plates. Assays were then performed according to the manufacturers' protocols. Absorbance was measured using a VersaMax Microplate Reader, Molecular Devices (Sunnyvale, CA, USA).

Statistical analysis

Statistical analysis was performed with Prism software version 6·02 (GraphPad Software, Inc., San Diego, CA, USA). The Mann–Whitney non‐parametric test was used to perform group comparisons. Correlations were calculated using Spearman's rank correlation. Data are expressed as mean ± standard deviation (s.d.) or median with the interquartile range (IQR), as appropriate. In all comparisons and correlations, a P‐value < 0·05 was used to indicate a statistically significant result.

Results

Subject characteristics

The clinical characteristics of study subjects with T1D and matched controls are shown in Table 1. Subjects with T1D were divided into two cohorts according to disease duration. For the purposes of this study, subjects with diabetes for less than 3 years were considered recent‐onset (RO) and the remainder were considered long‐standing (LS). While the overall T1D and control donors were matched closely for gender, the RO T1D subgroup have a higher frequency of males than the LS T1D subgroup. No significant differences in HbA1c were found between the two subgroups [RO T1D: 6·9 (6·4–8·3)% versus LS T1D: 7·1 (6·1–8·2)%; median (IQR)].

Table 1.

Characteristics of subjects in the study

	Controls	All T1D	RO T1D	LS T1D
Subjects	38	44	27	17
T1D duration (years) (1)	–	3·1 ± 3·1	1·1 ± 0·9	6·2 ± 2·6
Age (years) (1)	28·8 ± 7·3	28·2 ± 8·3	27·6 ± 8·4	29·1 ± 8·3
Gender
Male	21	25	21	4
Female	17	19	6	13
Ethnicity
White	33	34	20	14
Asian	1	7	5	2
Other	4	3	2	1
HbA1c (%) (2)	–	6·9 (6·3–8·3)	6·9 (6·4–8·3)	7·1 (6·1–8·2)
HbA1c (mmol) (2)	–	52 (45–67)	52 (46–67)	54 (43–66)

Data are expressed as either (1) mean ± standard deviation or (2) median and interquartile range. T1D = type 1 diabetes; RO = recent‐onset; LS = long‐standing.

Serum levels of NETosis‐associated biomarkers

The most marked elevation in a serum NETosis‐associated biomarker reported previously was in PR3 9, so we first evaluated levels of this enzyme in our cohorts using a commercial ELISA. In contrast to previous findings, average serum PR3 levels were actually modestly decreased, by 26·2%, in subjects with T1D compared with matched controls (Fig. 1a). The reduction in PR3 levels was most pronounced in RO T1D subjects, which showed less circulating PR3 than both controls and the long‐standing T1D group, suggesting that the finding could be related to disease duration (Fig. 1b).

Figure 1.

Circulating proteinase 3 (PR3) and neutrophil elastase (NE) levels are reduced in subjects with recent‐onset type 1 diabetes (T1D). Levels of PR3 and NE in all T1D subjects versus matched healthy controls (HC) (a,c) or in T1D subgroups separated by disease duration (b,d) are shown. Subjects were considered recent‐onset (RO) if diagnosed within 3 years or long‐standing (LS) if diagnosed greater than 3 years prior. Statistical significance is indicated by asterisks (*P < 0·05; **P < 0·01, Mann–Whitney U‐test). Horizontal lines indicate median values.

To address whether this reduction could be observed with other NETosis markers, we proceeded to analyse NE, a second marker that was reported to be elevated in T1D along with PR3 9. As with PR3, NE was also found at lower average levels in T1D, with the lowest levels being found in the subgroup closer to diagnosis (Fig. 1c,d). The agreement of these independent assays strengthens the likelihood that there were real biological reductions in NETosis‐associated markers in our T1D cohort.

Association of NETosis markers with neutrophil count

We were surprised to find that NETosis proteins were reduced in T1D, which was the opposite of what was expected. However, as these markers are products of neutrophils, we thought the findings could be related to a reduction in neutrophil numbers, which has been reported previously by multiple groups 7, 8. Consistent with these reports, the RO T1D subgroup had reduced numbers of neutrophils compared to controls (3·2 ± 0·9 × 10³/μl versus 4·4 ± 1·8 × 10³/μl; mean ± s.d.) and T1D subjects with longer disease duration (3·2 ± 0·9 × 10³/μl versus 4·8 ± 2·0 × 10³/μl) (Fig. 2a). Indeed, the PR3 and NE levels were correlated with neutrophil count (Fig. 2b,d). Finally, when the serum NETosis marker levels were normalized to neutrophil count for individual subjects, there was no longer a difference between T1D subjects of either duration with controls (Fig. 2c,e). Therefore, we conclude that circulating PR3 and NE in the populations studied are largely reflections of absolute neutrophil counts.

Figure 2.

Serum proteinase 3 (PR3) and neutrophil elastase (NE) levels reflect mild neutropenia in recent‐onset type 1 diabetes (T1D) subjects. For a subset of subjects with available complete blood count (CBC) results, the absolute number of neutrophils per mL blood was determined. The absolute neutrophil count (ANC) for control subjects and subjects with recent‐onset (RO) or long‐standing (LS) T1D are shown (a). Statistical significance is indicated by asterisks (*P < 0·05; Mann–Whitney U‐test). The correlation of PR3 and NE levels with ANC was analysed by Spearman's rank correlation (b,d). The serum concentration (pg/mL) of PR3 and NE normalized to blood neutrophil concentration (cells/mL) is shown (c,e). Normalized levels of NETosis‐associated markers did not differ significantly between groups. Horizontal lines indicate median values.

There was no evidence of an effect of ethnicity on PR3 and NE levels when comparing Asian and Caucasian T1D subjects (Supporting information, Fig. S1A,E), nor were they affected by gender (data not shown). PR3 showed weak correlation with age (Supporting information, Fig. S1B) and glutamic acid decarboxylase (GAD65) autoantibody levels (Supporting information, Fig. S1D), although these associations were not significant for NE (Supporting information, Fig. S1F,H). Neither marker correlated with other clinical characteristics, such as other autoantibodies or HbA1c (Supporting information, Table S1).

Discussion

Our analysis of adult T1D subjects in North America indicates that the concentration of circulating NE and PR3 are significantly, although modestly, reduced in the circulation in concert with reduced absolute neutrophil counts. These results are in contrast to a recent study of paediatric subjects with T1D in China 9, and although the reasons for the discrepancy are not clear, they could be related to the populations studied (onset age, genetics, environment, standard of care) or to the analytical methods employed. We did not see any indication that ethnicity was a clear explanation, although the number of subjects of Chinese descent in our study was limited and a relevant difference in genetics between the populations cannot be ruled out. We observed a weak correlation between age and PR3 levels, but this association was in the positive direction, with the younger paediatric cases (aged 15 years) in our cohort showing lower PR3 levels than adults. This result is not unexpected, as chronic up‐regulation of inflammatory mediators are associated with ageing 17. Extrapolating from these data, we do not expect that younger T1D cases in our population are likely to display elevation in NETosis markers, although it cannot be ruled out, particularly for subjects with early‐onset age and short disease duration. Recent reports have shown that neutrophils cultured from types 1 and 2 diabetic patients are more prone to undergo NETosis in vitro, and this increased susceptibility was related to elevated glucose 18, 19. Notably, our study was conducted on blood samples collected without fasting, yet still failed to show evidence of systemic NETosis induction that might be blood glucose‐related. Further, NE and PR3 levels were not related to HbA1c, excluding probably strong contributions of persistent hyperglycaemia or associated metabolic abnormalities that might differ between the populations studied. Although directly examining neutrophils from subjects with T1D during acute inflammation was beyond the scope of this study, our results suggest that during steady state, neutrophils in T1D subjects do not increase NET production due to typical hyperglycaemia. However, it is important to stress that this does not exclude the possibility that transiently increased NETosis occurs in diabetic subjects during inflammatory events, in localized tissues such as the islets and/or during more severe hyperglycaemic episodes. We also cannot rule out differences in NE or PR3 enzymatic activity levels that could possibly be associated with disease status or glucose levels. In either case, we do not see any evidence that system‐wide NETosis activity, as assessed in the blood, is a regular feature of T1D in adults.

In contrast to our original hypothesis, the primary finding was that of reduced circulating PR3 and NE levels in those with recently diagnosed T1D, reflecting a mild neutropenia occurring consistent with previous reports 7, 8. The reason for these neutrophil reductions is still not clear. Based on the similarity in NE and PR3 levels in healthy and T1D subjects when corrected for circulating neutrophil numbers, there is no evidence to suggest that neutrophils from T1D patients are globally more prone to undergo NETosis, although this question would be better addressed by studying freshly isolated cells ex vivo. However, there are several potential explanations for the observed neutropenia, including death via other pathways, reduction in neutrophil myelopoeisis in the bone marrow or depletion from the circulation due to accumulation in inflamed islets.

Given the increasing evidence of the role of neutrophils in autoimmune diseases including T1D, it will be important to continue to study the relationship of neutropenia around the time of T1D onset with clinical outcomes such as the rate of decline in beta cell function. One challenge is that, in many retrospective clinical studies from non‐clinical trial settings, reliable complete blood count data is not always available. Our data suggest that circulating neutrophil products, such as PR3, NE and/or related markers, could potentially be developed to serve as serum‐based surrogates for absolute neutrophil count.

Disclosure

The authors declare no financial or commercial disclosures related to this study.

Author contributions

J. Q., S. F., C. S. and J. M. O. designed the experiments, C. J. G. oversaw the human subjects research, J. Q. and S. F. performed the experiments and analysed the data and J. Q. and J. M. O. wrote the manuscript. Funding for this project was provided by the NIDDK Medical Student Research Program in Diabetes and Obesity (to S. F.) and by internal BRI laboratory funds (to J. M. O.).

Supporting information

Additional Supporting information may be found in the online version of this article at the publisher's web‐site:

Fig. S1. Effects of race (a,e), age (b,f), HbA1c (c,g) and glutamic acid decarboxylase (GAD65) autoantibody (d,h) of type 1 diabetes (T1D) subjects on serum concentrations of proteinase 3 (PR3) and neutrophil elastase (NE). P‐values were calculated using the Mann–Whitney U‐test, horizontal line represents mean value (a,e). Spearman's rank correlation was used to calculate coefficient of correlation r_s (b–d, f–h).

Table S1. Spearman's rank correlations of circulating neutrophil elastase (NE) and proteinase 3 (PR3) levels with clinical parameters in type 1 diabetes (T1D) patients studied.