Glycation of CD59 impairs complement regulation on erythrocytes from diabetic subjects

Abstract

Type 1 diabetes is associated with anaemia. Although the underlying mechanisms remain unclear, the accompanying reticulocytosis implies that erythrocyte lifespan in the circulation is shortened. Among the factors that permit prolonged survival of erythrocytes are the membrane complement regulators. In conditions such as paroxysmal nocturnal haemoglobinuria, where erythrocyte expression of these regulators is reduced, erythrocyte survival is compromised and anaemia follows. Recent in vitro evidence indicates that one of the key membrane complement regulators, CD59, is inactivated by glycation in the presence of high concentrations of glucose or other glycating sugars. To ascertain whether glycation-induced inactivation of CD59 occurrs in vivo we examined CD59 surface expression and function on erythrocytes from a cohort with poorly controlled type 1 diabetes (hyperglycaemic) and from matched normoglycaemic controls. Although expression of CD59, assessed using polyclonal anti-CD59 antiserum, was similar in the two groups, erythrocytes from hyperglycaemic individuals were more susceptible to lysis by complement, entirely as a result of the loss of functional CD59. These data implicate glycation-induced inactivation of CD59 as a factor contributing to anaemia in type 1 diabetes.

Introduction

Diabetes mellitus is one of the most common metabolic disorders, encompassing a range of conditions associated with hyperglycaemia. Individuals with diabetes are at increased risk of long-term complications such as retinopathy, nephropathy, neuropathy and macrovascular disease affecting the cardiovascular, cerebrovascular and peripheral vascular systems.¹ Diabetic microvascular disease is strongly associated with poor glycaemic control.^2,3 Macrovascular disease is the result of complex heterogeneous interactions in which glycaemic control may play a role.⁴ Patients with type 1 diabetes are also at increased risk of developing anaemia, this is probably related to the reduced erythrocyte (E) half-life that is commonly observed in diabetic subjects.^5,6

Mechanisms that have been implicated in the development of complications of diabetes include glycation-mediated protein inactivation and the generation of advanced glycation end-products. Glycation is the process through which reducing sugars react with amino groups that are located close to imidazole moieties or that are part of a lysine doublet, the so-called glycation motif, in proteins.⁷ As a consequence of the addition of the sugar moieties, the function of the protein may be impaired.⁸ Numerous proteins are glycated in vivo in diabetics with poor glycaemic control, although in many cases the functional consequences of glycation are unclear. For immunoglobulin G, glycation in the Fc portion of the molecule prevents effector recruitment, rendering the molecule inert.^9,10 Advanced glycation end-products are the final products of protein glycation and oxidation which bind specific receptors on endothelia and other cell types to amplify inflammatory responses.¹¹

The complement (C) system is an essential component of innate immune defence, providing protection from invading organisms and a mechanism to deal with immune complexes.¹² It comprises some 14 plasma proteins, together with a larger number of regulatory proteins, present both in plasma and on cell membranes that prevent unwanted activation. Glycation in vivo of C components in diabetic subjects has been described,^13–15 although the effects of glycation on function were not explored. Functional inactivation of C components by glycation might contribute to the observed increased susceptibility of diabetics to bacterial infections. The effect of poor glycaemic control on C regulators has been little explored. Glycation of the fluid-phase regulator vitronectin has been described but the effects on C regulatory function were not tested.¹⁶ Three membrane proteins, CD46, CD55 and CD59, collaborate to protect self cells from C, the first two acting as inhibitors during C activation and the last, CD59, acting on the terminal stage of the C pathway to regulate assembly of the membrane attack complex.¹⁷ Insertion of the membrane attack complex into the membrane of a target cell creates a pore, thereby causing osmotic lysis. Human CD59 contains a glycation motif at K41.¹⁸ It has been shown in vitro that incubation of CD59 in the presence of glycating sugars causes a loss of C regulatory function.¹⁹ Preliminary analyses from this same study suggested that CD59 isolated from diabetic urine was glycated. Together, these findings provoked the suggestion that glycation of CD59 on plasma-exposed cells in diabetic subjects might render the cells susceptible to damage by C.

On the majority of blood cell types and endothelia, CD59 is turned over relatively rapidly, making it unlikely that glycation, a slow process even in the presence of very high concentrations of glycating sugars, would have significant functional consequences. The exception is the E, a long-lived (120 days) cell abundantly expressing CD59 and with no turnover of surface proteins. We therefore set out to examine the effects of poor glycaemic control on the expression and function of CD59 in diabetic E. We first confirmed the observation that CD59 was susceptible in vitro to functional inactivation when incubated with glycating sugars and then examined CD59 expression and C inhibitory function in E from poorly controlled diabetics and matched controls. The results show a remarkable loss of CD59 function that renders diabetic E susceptible to lysis by homologous C.

Materials and methods

Patient samples

For studies of lytic susceptibility, E from 20 poorly controlled type 1 diabetic patients (11 male, nine female; mean age 62·2 ± 14·4 years), and 20 age-matched non-diabetic controls (11 male, nine female; mean age 61·7 ± 15·9 years) were investigated. All diabetic individuals had essentially normal renal function as assessed by measurement of serum creatinine. For studies of CD59 surface expression, this group was expanded to 40 diabetic (23 male, 17 female; mean age 67·2 ± 12·6 years) and 40 non-diabetic subjects (23 male, 17 female; mean age 67·9 ± 13·4 years). Peripheral blood was collected into vacutainers containing ethylenediaminetetraacetic acid (Becton Dickinson, Franklin Lakes, NJ) from diabetic and non-diabetic subjects with informed consent. Samples were diluted 50 : 50 (v/v) in Alsever's solution (114 mm sodium citrate, 27 mm glucose, 72 mm NaCl, pH 6·1) and stored at 4° for up to 7 days. For all diabetic subjects, glycaemic control was assessed in the same samples by measurement of HbA1c levels using reverse-phase cation-exchange chromatography (A. Menarini Diagnostics, Oxford, UK). All non-diabetic control samples were confirmed normoglycaemic and HbA1c levels were measured.

Materials

All chemical reagents, unless otherwise stated, were purchased from Sigma Chemical Company (Poole, Dorset, UK). Guinea-pig erythrocytes (GPE) were obtained from the university animal facility. Veronal-buffered saline (VBS; 2·8 mm barbituric acid, 145·5 mm NaCl, 0·8 mm MgCl₂, 0·3 mm CaCl₂, 0·9 mm sodium barbital. pH 7·4) was obtained from Oxoid Ltd (Basingstoke, UK).

Antibodies

Monoclonal antibodies (mAbs) were obtained from the following sources: BRIC 229 (anti-CD59) from International Blood Group References Laboratory (IBGRL; Bristol, UK). The hybridomas YTH53.1 and MEM43 (both anti-CD59) were purified, as previously described²⁰ from supernatants of cell lines obtained from H. Waldmann (Oxford, UK) and V. Horejsi (Prague, Czech Republic), respectively. Rabbit polyclonal anti-CD59 and anti-CD55 were prepared in house by immunization with proteins purified from E membranes. Secondary antibodies were obtained from DAKO (Cambridge, UK).

In vitro glycation

Native CD59 was isolated from butanol extracts of human E as described.²¹ CD59 (20 μg) was incubated for 30 days at 37° in the presence of 500 mm of the glycating sugar d-Ribose, 500 mm of the non-glycating sugar sorbitol or phosphate-buffered saline (PBS; 8·1 mm NaH₂PO₄, 137 mm NaCl, 2·7 mm KCl, pH 7·4). Samples were taken every 5 days for assessment of CD59 function, drop-dialysed against VBS (30 min, 4°) and stored frozen until assayed.

CD59 function was assessed using a C protection assay. GPE were washed three times in PBS, harvested by centrifugation at 2000 g for 10 min and resuspended to a 5% (v/v) suspension in PBS. CD59 (1 μg) was added to 200 μl 5% (v/v) GPE and incubated at 37° for 30 min to allow CD59 to incorporate into the GPE cell membrane. Unincorporated CD59 was removed by washing cells three times in PBS and the cells were resuspended to the original volume in VBS. Incorporation was confirmed by staining with polyclonal anti-CD59 and flow cytometry as described below. GPE were then incubated with an equal volume of a 1 : 5 (v/v in VBS) dilution of human AB serum at 37° for 30 min. This serum dilution was selected in preliminary experiments to give just submaximal lysis of untreated GPE. The cells were harvested by centrifugation at 2000 g for 10 min. The absorbance of the supernatant at 415 nm (Abs_sample) was recorded using a Bio-Rad Microplate reader (Model 3550-UV), and corrected for background absorbance (Abs_serum). The absorbance reading, relative to that obtained for complete lysis of E, achieved by dilution of 5% E (50 : 50 v/v) in water (Abs_100%) was used to calculate percentage cell lysis according to the following formula: % cell lysis = [(Abs_sample − Abs_serum)/(Abs_100% − Abs_serum)] × 100.

Haemolytic susceptibility assays

E from normal and diabetic subjects were washed three times in PBS, harvested by centrifugation at 2000 g for 10 min and resuspended to 2% (v/v) in VBS. Polyclonal anti-CD55 antibody was serially diluted and incubated with 50 μl 2% (v/v) E to determine the maximum titre of anti-CD55 that did not cause agglutination of E. This amount of anti-CD55 polyclonal antibody was then used to sensitize the E for complement lysis and concomitantly neutralize CD55 on the E. Binding of anti-CD55 to E was confirmed by flow cytometry and levels of binding on normal and diabetic E were compared. Aliquots of antibody-sensitized E were placed in triplicate into a 96-well plate. Triplicate wells were incubated with either VBS alone, a 1 : 2 dilution of human serum (blood group AB) or a 1 : 2 dilution of human serum plus blocking antibodies to CD59 (BRIC 229 and MEM43, each at 10 μg/ml). The plate was incubated at 37° for 30 min, after which it was centrifuged at 800 g for 3 min. The supernatants were transferred into a fresh flat-bottom 96-well plate and their absorbance at 415 nm was measured.

CD59 cell surface expression

The expression of CD59 on the surface of normal and diabetic E was assessed using three different mAb and the polyclonal antibody. E from normal and diabetic subjects were washed three times in PBS, harvested by centrifugation at 2000 g for 10 min and resuspended to a 0·1% (v/v) solution in flow cytometry buffer [PBS supplemented with 1% bovine serum albumin (w/v), 0·2% NaN₃ (w/v), pH 7·4]. Aliquots (100 μl) were incubated with 1 μg of the appropriate anti-CD59 antibody at 4° for 30 min. Antibody binding was detected using fluorescein isothiocyanate-conjugated rabbit anti-mouse immunoglobulin (5 μg/ml) or phycoerythrin-conjugated goat anti-rabbit immunoglobulin (10 μg/ml). Fluorescence was measured using a Becton Dickinson FACScalibur.

Statistical analysis

The significance of differences between groups was assessed using an unpaired Student's t-test (two-tailed, Welch-corrected). Differences were considered significant when P < 0·05.

Results

Glycation in vitro reduces the C regulatory activity of CD59

CD59 purified from human E retains its glycosylphosphatidylinositol (GPI) anchor and therefore can be attached to the membranes of target cells for assessment of function. CD59 was incubated with the glycating sugar, d-ribose, or the non-glycating sugar, sorbitol, for intervals up to 30 days. Others have used a similar protocol to glycate CD59 in vitro and confirmed glycation using a specific antibody against the glycated protein.¹⁹ We were unable to obtain this reagent to formally confirm glycation but can confidently make this assumption based on the similarities in protocols used. Incubation with sugars did not alter the migration properties of CD59 on sodium dodecyl sulphate–polyacrylamide gel electrophoresis (not shown). To assess C regulatory activity, CD59 was then incorporated into GPE. Incubation with sugars did not affect CD59 incorporation into GPE as confirmed by staining with polyclonal anti-CD59 and flow cytometric analysis (data not shown), further demonstrating that the integrity of the protein was not compromised by the treatment. Whereas CD59 preincubated for up to 30 days in PBS alone or in non-glycating sugar was protective against C lysis when incorporated into GPE, exposure of CD59 to the glycating sugar d-ribose prior to incorporation caused a marked reduction in its ability to protect from C lysis, first apparent after 15 days of incubation at 37°(Fig. 1). By day 30, C protective activity was almost completely lost when compared to untreated GPE.

Figure 1.

The effect of glycation upon C regulatory function of CD59 in vitro. CD59 was incubated at 37° in the presence of the glycating sugar d-ribose, the non-glycating sugar sorbitol or buffer control (PBS). Samples were removed at intervals and stored frozen. At the end of the experiment, all samples were assayed for C regulatory activity in a single assay as described in the Materials and methods section. Percentage inhibition of lysis was calculated for each sample. Results are means of triplicate determinations and bars represent SD of triplicates.

Diabetic E display increased susceptibility to lysis by C

The lytic susceptibility of E from 20 poorly controlled diabetics [HbA1c 9·5% ± 1·2 (mean ± SD)] and 20 age-matched non-diabetic controls (HbA1c 5·7% ± 0·7) was investigated. In the absence of sensitizing antibody, neither normal nor diabetic E were lysed by human serum. E were sensitized and decay-accelerating factor (DAF) was blocked by incubating with a polyclonal anti-DAF antiserum that bound equally to diabetic and non-diabetic E, as assessed using flow cytometry (data not shown). The serum dose used was chosen to give a consistent, low degree of lysis of non-diabetic E incubated with the anti-DAF antiserum. E from diabetic patients displayed markedly increased lysis following exposure to C attack compared to E from non-diabetic subjects. The mean level of lysis in the non-diabetic samples was 5·3 ± 3·23% (SD; n = 20), whereas in the diabetic samples mean lysis was 55·1 ± 16·1% (n = 20), a statistically highly significant difference (P < 0·0001) (Fig. 2). To confirm that this difference was the result of functional inactivation of CD59, E from diabetic and non-diabetic individuals, sensitized with anti-DAF antiserum, were preincubated with function-blocking anti-CD59 mAb prior to exposure to C. Blockade of CD59 markedly enhanced C lysis of non-diabetic E [mean lysis ± SD: blocked 76·5 ± 15·3%; unblocked 5·3 ± 3·23%, P < 0·001), but caused a small, albeit significant, increase in lysis of diabetic E (blocked 77·7 ± 12·1%; unblocked 55·1 ± 16·1%, P < 0·001). In the presence of blocking mAb against CD59, the lysis of non-diabetic and diabetic E was equivalent (diabetics: 77·7 ± 12·1%, n = 20; non-diabetics: 76·5 ± 15·3%, n = 20).

Figure 2.

Haemolytic susceptibility of diabetic and non-diabetic erythrocytes (E). Washed E from normal and diabetic subjects (20 in each group) were sensitized with polyclonal anti-CD55 antibody. Antibody-sensitised E were then incubated with either VBS alone, human serum (blood group AB) or human serum plus blocking antibodies to CD59 as described in the Materials and methods section. Percentage lysis was measured. Bars represent means ± SD for each experimental group. The significances of difference are detailed in the text.

Diabetic E retain surface expression of CD59 but lose epitopes for active site-specific mAb

E from 40 diabetic subjects [HbA1c 9·9 ± 1·6% (mean ± SD)] and 40 non-diabetic subjects (HbA1c 5·5 ± 0·6%) were stained with each of three anti-CD59 mAbs (MEM43, BRIC229 and YTH53.1) and one polyclonal anti-CD59 (poly-CD59) and analysed in triplicate by flow cytometry. CD59 expression was determined by measurement of the mean intensity of fluorescence on each sample. Forward and side scattering profiles were similar for diabetic and non-diabetic E. With each of the mAb, staining for CD59 was significantly lower on diabetic E than on matched non-diabetic controls (Table 1; Fig. 3). However, when CD59 expression was analysed using a specific polyclonal antibody, expression levels of CD59 on diabetic and non-diabetic E were similar (Table 1; Fig. 3).

Table 1. Anti-CD59 antibody staining of diabetic and non-diabetic erythrocytes.

	Fluorescence (mean ± SD)
Antibody	Diabetic	Non-diabetic	P-value
YTH53.1 mAb	93·9 ± 19·8	125·4 ± 18·4	< 0·001
BRIC229 mAb	606·9 ± 57·7	804·1 ± 87	< 0·001
MEM43 mAb	260·5 ± 49·4	314·4 ± 39·7	< 0·005
Polyclonal anti-CD59	93·3 ± 8·6	95·2 ± 9·4	Not significant

Figure 3.

Flow cytometric analysis of CD59 expression on diabetic and non-diabetic erythrocytes (E). Washed E were stained with anti-CD59 mAb MEM43 (a), BRIC229 (b), YTH53.1 (c) or polyclonal anti-CD59 (d) followed by appropriate fluorescein isothiocyanate-labelled or phycoerythrin-labelled secondary antibodies as detailed in the text. The figure shows typical examples of profiles obtained with diabetic E (D) and normal E (ND). Isotype control staining (C) for the same samples is shown in each case and was identical for diabetic and non-diabetic E.

Discussion

Anaemia is a well-recognized feature of Type 1 diabetes, and has variously been ascribed to erythropoietin deficiency or insensitivity and to associated renal impairment.^5,22–24 However, anaemia is seen even in the presence of normal renal function, the lifespan of E in diabetic patients is reduced by as much as a third compared to normal individuals and the anaemia is compensated by a mild to moderate reticulocytosis, indicating increased erythropoietic activity.⁵ These observations suggest that other factors, directly influencing E survival, might play an important role in diabetic anaemia.

The C cascade, an important defence against pathogens, has the capacity also to damage or destroy self-cells and expression of the membrane C regulators, CD46, CD55 and CD59, is essential to survival of self-cells in vivo. The haemolytic and thrombotic disorder paroxysmal nocturnal haemoglobinuria (PNH) is characterized by the absence of GPI-anchored molecules, including the C regulators CD55 and CD59, and manifests with spontaneous haemolysis and platelet dysfunction as a result of failed C regulation.²⁵ Human E do not express the third membrane C regulator CD46, and CD35, which is expressed in small amounts on human E, acts exogenously and does not protect the expressing cell; hence, E in PNH patients are effectively devoid of membrane C regulators. CD59 is the sole cell membrane regulator of the terminal pathway. It acts to restrict the formation of the membrane attack complex, thereby preventing inappropriate damage to target cells. Absence or loss of function of CD59 has the potential to cause widespread cell damage. Absence of CD59 is the critical deficit in PNH, as is apparent from studies of individuals who are genetically deficient in CD55 or CD59 – only the latter have a PNH-like disease.^26,27 Further confirmation of the key role of CD59 in protecting E and other blood cells was provided by the creation of CD59-knockout mice and the demonstration that these mice had PNH-like symptoms.²⁸ Loss of the protection afforded by CD59 on E might therefore be anticipated to cause haemolysis.

The possibility that CD59 might be compromised in diabetics was suggested by the demonstration that human CD59 contained a glycation motif centred around amino acid residue K41 in close proximity to W40, a highly conserved amino acid that is known to be essential for CD59 function.^18,29 The close proximity of the glycation motif to W40 suggested that glycation might affect the activity of CD59. Indeed, loss of function of purified CD59 following prolonged incubation with glycating sugars has been described.¹⁹ These workers also demonstrated that human E incubated in vitro with high concentrations of glycating sugars became more susceptible to lysis by C. Here we have confirmed that CD59 is slowly inactivated when incubated with glycating sugars in vitro. Glycated CD59 still incorporated into target cells but did not confer protection from C lysis. We have now extended these observations to the in vivo situation, demonstrating that E from poorly controlled diabetic patients, when sensitized with an antiserum that simultaneously activates C and blocks DAF function, show markedly increased susceptibility to C lysis when compared with non-diabetic E. Diabetic control here was assessed by measuring HbA1c, a good index of glycaemic control, over the preceding few weeks; measurement of plasma glucose at the time of clinic attendance is a poor index of control. To discover whether CD59 was lost from diabetic E, antibody staining and analysis by flow cytometry were performed. Staining with each of three different mAb against CD59 suggested a marked loss of CD59 protein from the cell surface when compared to normal E, but staining with a specific polyclonal antibody was similar for diabetic and non-diabetic E. All three of the mAb used were functional blockers of CD59 and their epitopes have been mapped to regions close to the putative active site of the molecule centred on W40 and immediately adjacent the glycation site at K41.¹⁸ Single residue mutations in this region have previously been shown to cause loss of binding of these same mAb while having little effect on binding of a polyclonal antiserum.¹⁸ These data indicate that all three mAb bind closely related epitopes, all of which are modified upon glycation, and that expression of CD59 on diabetic E is not significantly altered. Blocking of CD59 expressed on normal E with mAb increased lytic susceptibility to a level comparable with that of untreated diabetic E, whereas the blocking mAb caused only a small further enhancement of lysis of diabetic E, presumably by blocking residual unglycated CD59 on the ‘youngest’ E. These observations confirm that loss of CD59 function was responsible for the increased susceptibility to lysis and show that diabetic E have little residual active unglycated CD59.

Diabetic E were not lysed by human serum in the absence of the anti-DAF sensitizing antibody. This mirrors the situation in PNH where E are lysed only when activation is induced either using sensitizing antibody or by serum acidification in the Ham test. Indeed, our preliminary data indicate that diabetic E are lysed in the Ham test. The ongoing tickover C activation in vivo will thus cause a low degree of haemolysis of the unprotected E in PNH and in diabetes through similar mechanisms.

The relevance of this observation may not be restricted to diabetes. Blood for use in transfusions is stored in a medium containing d-glucose at 35 mm for up to 35 days at 4°. Storage of blood in this manner has been shown previously to increase the proportion of glycated haemoglobin³⁰ and it is therefore likely that CD59 on stored E is to some degree glycated and inactivated, even with storage at low temperature. Transfused E may then be more susceptible to C lysis and display a shortened lifespan in vivo. In this study, E were stored in Alsever's solution containing glucose at 27 mm for up to 7 days. This relatively brief exposure did not alter the susceptibility of the stored E to lysis; nevertheless, further studies of the effects of storage media on the function of E CD59 will be informative.

Whether CD59 on other plasma-exposed cell types in diabetic individuals is similarly compromised remains to be tested. E are a special case in that they must survive in the circulation for several months and do not have the capacity for membrane protein turnover. In contrast, CD59 on leucocytes and endothelia will, in common with other GPI-anchored molecules, be turned over rapidly, perhaps too quickly for significant glycation of CD59 to occur. Platelets, while having only a limited capacity to turnover membrane proteins, are short-lived and are thus also likely to escape the major effects of glycation, although this should be tested experimentally. An intriguing report recently described a second mechanism for loss of C regulators in the face of hyperglycaemia.³¹ Cultured endothelial cells exposed to high d-glucose concentrations showed reduced surface expression of both CD55 and CD59, apparently through shedding following activation of phosphatidyl inositol-specific phospholipase C. High plasma glucose may therefore act in different ways on different cell types to reduce the expression or function of CD59 and render cells susceptible to C damage or destruction.

Abbreviations

C

complement

DAF

decay-accelerating factor

E

erythrocyte

GPE

guinea-pig erythrocytes

GPI

glycosylphosphatidylinositol

mAb

monoclonal antibody

PBS

phosphate-buffered saline

PNH

paroxysmal nocturnal haemoglobinuria

VBS

veronal-buffered saline