This invention relates to methods for detecting autoimmune diseases and pre-clinical autoimmune diseases.

In particular, it relates to methods for detecting insulin dependent diabetes mellitus (IDDM) and pre-clinical IDDM. Peptide fragments are provided for use in these methods.

Background

Epidemiological evidence in man (1-4) and data from animal feeding studies (5-8) have suggested a diabetogenic effect of dietary cow milk proteins. Supportive serological findings have been identified in animals (5,9) and humans (10-12) associating immunity to cow\'s milk proteins and Type 1 diabetes. The most direct evidence for a pathogenic link between cow\'s milk proteins and diabetes comes from a family study in Finland, where exclusive breast-feeding for the first 3-4 months of life was found to protect from later development of diabetes (4) .

Most of these studies did not identify a specific cow milk protein or explain the near global increase in diabetes incidence despite emphasis on breast-feeding. However, these latter observations (4) are consistent with the view that in humans (as in diabetes-prone rats (13) a diabetes associated immune response to bovine serum albumin (BSA) or to peptide fragments or portions thereof is triggered in the early post-natal period (14,15).

An existing detection method for pre-clinical IDDM is based on detection of islet cell antibodies

(16-18) . Unfortunately, this is a very difficult and laborious (2-day) assay which requires manual processing, visual judgment of results with its inherent inaccuracies. There are only a few centers in any country which can perform this assay according to international standards. Up to 20% of the population and 30-50% of IDDM family members have such antibodies. The test predicts clinical IDDM in only a subset of cases and is positive at diagnosis in only 80%.

Animal work with BB diabetic rats showed detectable antibodies to bovine serum albumin in IDDM and in some animals without IDDM but with some histological islet changes (13) . Since not all animals showing such changes progress to overt IDDM, these studies did not indicate that detection of antibodies to bovine serum albumin could be predictive of progression to IDDM.

Until the work of the present inventors, no convenient clinical method with sufficient predictive and discriminatory ability was available to screen human subjects and detect pre-clinical IDDM.

Figures Figure 1: 1A: Serum IgG anti-BSA Antibody Concentrations at Diagnosis of Insulin-Dependent Diabetes in 142 Children (diagonal hatching) and 79 Normal Children (horizontal hatching) . Figure IB shows distribution of serum IgA anti-BSA Antibody Concentrations and Figure 1C shows serum IgM anti-BSA Antibody Concentrations, both after normalisation by smoothing.

Figure 2: Measurements of total and ABBOS-Specific Anti- BSA Antibodies in 17 Diabetic Patients. Panel 2A shows IgG, panel 2B IgA and panel 2C IgM. The hatched bars represent patient anti-BSA levels, the white bars the

levels remaining after removal of ABBOS-specific antibodies. The horizontal bar indicates the mean anti- BSA concentrations in 17 normal children (upper line) and the concentrations after removal of anti-ABBOS antibodies in the same sera.

Figure 3: 3A shows distribution of moderately- (horizontal hatching) , highly (diagonal hatching) , and very highly elevated (vertical hatching) IgG-anti-bovine serum albumin antibodies in 40 diabetic children as determined by particle concentration fluoroimmunoassay (PCFIA) . 3B and 3D: PCFIA standard curve for a serum pool (from diabetic children) containing (3B) 12.3 KfU/μl IgG- and (3D) 4.2 KfU/μl IgA-anti-BSA antibodies: binding competition with increasing amounts of BSA and ovalbumin/Tween-20. 3C shows IgG anti-BSA standard curve for enzyme im unoassay (EIA) . 3E shows IgA anti-BSA standard curve for enzyme immunoassay (EIA) . 3F shows binding competition with increasing amounts of free ovalbumin/Tween-20 for IgG-(θ) and IgA-(___>) anti-BSA antibodies, as well as with increasing amounts of free BSA for IgG-(D) and IgA- (■) anti-BSA antibodies. KfU=kilo fluorescence units.

Figure 4: Mean levels (±SEM) of anti-BSA antibodies in Type 1 diabetic- and matched control children as detected by particle concentration fluoroimmunoassay (PCFIA, Figures 4A and 4B) and enzyme immunoassay (EIA, Figures 4C and 4D) . Difference between diabetic and control children: PCFIA:IgG, p <0.0001; IgA, p <0.001. EIA: IgA, p <0.01.

Figure 5 - Correlation between the levels of anti-BSA antibodies as determined by enzyme immunoassay (EIA) and particle concentration fluoroimmunoassay (PCFIA) in diabetic- and control children. Shaded areas represent BSA-antibody levels considered as "non-elevated":

(vertical band) negative for BSA antibodies by PCFIA, (horizontal band) negative for BSA antibodies by EIA) . 5A: IgG in diabetic children, n=40, r _s 0.28, p=0.09; 5B: IgA in diabetic children, n=40, r,=0/ll, p=0.48; 5C: IgG in control children n=179, r _s =0.02, p=1.0; 5D: IgA in control children, n=179, r _s =-0.05, p=1.0. Correlation coefficients were determined by Spearman\'s rank correlation.

DESCRIPTION OF THE INVENTION

In the description which follows, references are made to certain literature citations which are listed at the end of the specification.

The measurement of autoimmune disease- associated antibody levels specific for proteins or protein fragments derived from common dietary sources as well as measurement of T lymphocyte sensitization to such fragments by the methods in the invention provide a unique and new clinical and investigational tool for

1.) the diagnosis of early, pre-clinical disease as prerequisite for the development and use of disease delaying or preventative therapies; 2.) the differential diagnosis of autoimmune patients at first clinical presentation;

3.) the monitoring of disease course and effects of therapy. The present inventors were the first to show by a prospective study that by determining human serum levels of antibodies to bovine serum albumin or to certain natural or synthetic peptide fragments thereof by a method to be described, one can detect those individuals who will develop IDDM.

In accordance with one embodiment of the invention, a method is provided for detecting IDDM or

pre-clinical IDDM by measuring serum antibodies to bovine serum albumin (BSA) by a particle concentration fluoroimmunoassay (PCFIA) technique as described in Examples 1 and 3. The relevant antibodies are detected by their binding to particle-bound BSA.

Other methods have been used to detect anti-BSA antibodies in serum, for example ELISA techniques, as described in Example 2. It can be seen from example 2 that the antibodies detected by the ELISA method do not provide the discrimination required for reliable diagnosis of diabetes or pre-diabetes.

Particle concentration fluoroimmunoassay detected elevated IgG-anti-bovine serum albumin antibodies in all diabetic children, enzyme immunoassay in 25% (p <0.0001). Fluoroimmunoassay detected elevated levels in 2.2% and enzyme immunoassay in 10% of control children (p <0.002). Elevated IgA-anti-bovine serum albumin antibodies in patients were slightly more often detected by fluoroimmunoassay than by enzyme immunoassay, while in control children enzyme immunoassays detected elevated levels three times more often (p <0.01). Values measured in either assay showed overall no correlation in either patient (IgG: r _s =0.28; IgA: r _s =0.11) or control sera (IgG: r, 0.02; IgA: r _s =-0.05). Fluoroimmunoassay for IgG was 100% disease-sensitive (enzyme immunoassay: 25%, p <0.0001) and more disease-specific (IgG; p <0.02).

The inventors\' findings demonstrate that these assay techniques detected distinct subsets of anti-bovine serum albumin antibodies with little (IgG) or some (IgA) overlap. In fluoroimmunoassay procedures, antigen:antibody binding occurs within 1-2 min while hours are allowed in an enzyme immunoassay. Antibodies with high on-off binding rates typical for immune response following hyperimmunization are therefore

measured preferentially by particle concentration fluoroimmunoassay and it is these antibodies which appear to be associated with diabetes. These observations emphasize the need for epidemiological surveys to validate immunoassay procedures used for clinical purposes.

Comparison of the amino acid sequences of various serum albumin proteins, including human and bovine proteins, suggested to the inventors to focus on the region between a ino-acids 138-166, the region of greatest divergence between human and bovine (19) .

In accordance with a further embodiment of the invention, various novel peptides within this region have been synthesised, as described in Example 4.

It has been found that the bulk of the diagnostic antibodies detected by PCFIA as described above bind to peptide CS2185, ABBOS, as described in Example 1.

The PCFIA assay described above may be performed using particle-bound BSA-peptides such as particle-bound ABBOS instead of particle-bound BSA.

The ABBOS peptide can be used to detect up to 90% of IDDM-associated anti-BSA antibodies even when modified at the C-terminus as described in Example 4.

These peptides may be fragments of the natural BSA protein or synthetic peptides prepared by a suitable technique. Such techniques will be known to those

skilled in the art and include chemical synthesis and recombinant techniques.

Analogues of these peptides which retain their ability to bind to IDDM-associated anti-BSA antibodies are included within the scope of the invention.

In accordance with a further embodiment of the invention, a method is provided for detecting in IDDM patients and in pre-clinical IDDM subjects sensitised T lymphocytes which specifically recognise and proliferate in response to peptide fragments of BSA, including ABBOS and CS2267.

This detection method can be used to detect

IDDM, pre-clinical IDDM and also to detect reactivation of the β cell-attacking immune process in patients after jScell or islet transplantation.

In accordance with a further embodiment of the invention, a peptide is provided, CS2267, which binds specifically to the sensitised T lymphocytes present in IDDM and pre-clinical IDDM patients.

By coupling a peptide such as CS2267 to toxic compounds, immunotoxins can be prepared which are directed to and can destroy the specific T lymphocytes which mediate β cell destruction in IDDM, providing a means of arresting the disease process.

Example 1 - Detection of anti-BSA antibodies by PCFIA

Patient Populations: Blood samples were obtained from 142 Finnish children (83 males, 59 females, mean (±SD) age 8.4±4.3 years) with newly diagnosed insulin-dependent diabetes mellitus. Fifty patients had diabetic ketoacidosis, 48 diabetic ketosis only, and the remainder hypoglycemia alone. All were continuously dependent on at least one daily injection of human insulin and had increasing insulin dependence after diagnosis. We also studied 79 age-, sex- and region-matched children admitted for minor surgery (42 males, 37 females, mean age 8.413.1 yr) , and from 300 adult Toronto blood donors. Blood samples were obtained from the patients before the first insulin injection and 3 to 4 months later and, in a random subset of 44 patients, 1 to 2 years later. The serum samples from the two groups of children were sent coded to Toronto.

Clinical assessment included history and measurements of insulin and islet cell autoantibodies identified either by indirect immunofluorescence or complement fixation test . Sample volumes were insufficient for full titration of the earliest samples, islet cell antibody results are therefore expressed as positive or negative. The HLA-A, -B, -C, DR.Dw haplotypes of all patients were determined as described (17).

Measurement of anti-BSA Antibodies

Anti-BSA antibodies were measured by particle concentration fluoro-immuno assay (PCFIA ) as described in Dosch et al, (21) .

96 - well unidirectional flow vacuum filtration plates were used for the assay and phase separation procedures

were carried out by the robotic Screen Machine™ instrument (IDEXX, Portland, Me., U.S.A.) which is programmable for reagent additions, timed incubations, phase separations, washings and measurements of particle bound fluoresceinated secondary antibody, as described in Dosch et al, (22) and Cheung et al. (23) .

Two-hundred microlitres of BSA [Grade V, Sigma Chemical Co., St. Louis, Mo., U.S.A., 10% in phosphate- buffered saline (PBS; 40g NaCl, lg KCl, lg KH ₂ P0 ₄ , 5.75g Na ₂ HP0 ₄ , o.5g CaCl ₂ , 0.5g MgCl ₂ /5 litres distilled water, pH 7.2)] was coupled covalently (100 μl of 10 mg/ml 1- ethyl-3-(3-dimethylaminopropyl)-carbodimide) onto 400 μl (5% stock; IDEXX) carboxylated polystyrene beads (diameter 0.75 μm) . Subsequently, 10% Tween-20 in 1.0% ovalbumin-PBS was used as blocking agent. Concentrations down to 1% Tween-20 in 0.1% ovalbumin-PBS may be used as blocking agent. After repeated washings, beads were stored in 1% Tween-20-PBS. Over a period of 9 months the activity of the beads remained unchanged. Ovalbumin was obtained from Sigma.

Twenty microlitres of test serum dilutions (1:100-1:1,000) were added to microwells containing 20 μl of 1:20 diluted BSA-coated icrospheres (initial 2.5% weight/volume) . Up to ten plates were inserted into the Screen-Machine for programmed phase separations, washings and addition (100 ng/well) of affinity purified, custom BSA-free fluorescein-conjugated goat anti-human IgG, IgA,

and IgM (Fc-fragment specific, BioCan, Mississauga, Ontario, Canada) . Drying and precipitation of serum protein was avoided by short (1 min) incubation, phase separation and washing procedures at low (5mm Hg) vacuum pressure. Prior to reading, wells were vacuum dried for 1 min and fluroescence emission (472/512nm) was read under high vacuum from the concentrated particle cake at the bottom of the well. Kinetics of the system have been published (Dosch et al. (1988) above) . We have processed up to 3,000 replicate samples per day per instrument per operator.

A calibrated pool of serum from diabetic patents was used as the standard in each plate. This standard contained 12.3 kilofuorescence Units (KfU) IgG anti-BSA antibodies per microliter, 4.2KfU/μl IgA and 4.0 KfU/μl IgM anti-BSA antibodies. The anti-BSA assays had a sensitivity of 1.0(IgG, IgA) and 10.0 (IgM) ng/ml, the intraassay-and interassay coefficients of variation were 8.9% and 9.8%, respectively. Addition of BSA (but not ovalbumin) blocked antibody binding in a dose-dependent fashion. Anti-BSA antibody concentrations exceeding the mean plus 2 SD in the 79 normal children were defined as elevated.

Data Analysis: Antibody concentrations as expressed as kilofluorescence units per microliter relative to the standard serum pool. The results were analyzed using Chi-square statistics, parametric one-way analysis of variance, and Student\'s unpaired t-test for normally distributed values. The distribution of anti- BSA concentrations was normal for each isotype. In the

case of skewed distribution, the Mann-Whitney-U test and Spearman\'s rank-correlation test were used(r,) . Anti-BSA antibody concentrations after diagnosis were evaluated using a paired test.

Anti-BSA Antibodies: The serum IgG anti-BSA antibody concentrations in the diabetic patients were considerably higher than those in the normal children (Figure 1) , the mean concentration being almost seven fold higher (Table, 1,P <0.001, Table 2). The elevated serum IgG anti-BSA concentrations in the diabetic patients did not reflect generalized immune responses against nutritional antigens, since the patients and the normal children had similar serum concentrations of IgG antibodies to the major cow milk proteins casein and β- lactoglobulin (Table 1) .

The mean serum concentration of IgA anti-BSA antibodies was higher in the diabetic patients than in the normal children (P <0.0001, Table 1), but the values overlapped (Figure 1) . Two thirds of the diabetic patients (and two normal children) had elevated concentrations (P <0.0001, Table 2). Consistent with the young age of the diabetic patients, the patients who had IgA anti-BSA antibodies were older than those with low antibody concentrations (10.1±3.4 vs . 6.0±4.6 yr; P <0.0001) .

The mean serum concentration of IgM anti-BSA antibodies in the diabetic patients were slightly lower than those in the normal children (P <0.05, Table 1). Less than 1 percent of patients had elevanted

concentrations compared with 8 percent of normal children (P <0.01, Table 2, Figure 1).

No diabetic patient or normal child had IgD anti-BSA antibodies, and in a random subset of patients no IgE anti-BSA antibodies were detected.

Table 1. And cow milk antibody levels (KfU/μi±SEM) at diagnosis of Type 1 diabetes.

Table 2. Proportion of sent (%) with elevated anti-BSA antibody levels.

Table 3. Decline of anti-BSA-antibody levels (KfUAUlSEM) and proportions (mean) of ABBOS-specific anti-BSA antibodies followin diagnosis of Type 1 diabetes.

* N=44 n.d.» not determined

^Δ p <0.001. *p <0.0001

While low concentrations of (mainly IgM) anti- BSA antibodies were detected both in the diabetic and normal children, high concentrations of IgG and IgA anti- BSA antibodies were found only in the diabetic children. The latter findings indicated the existence of a close (100%) association between BSA-specific immune responses and the clinical expression of insulin-dependent diabetes. Consistent with an active, antigen-driven immune response against BSA in diabetes, there was a significant correlation between IgM and IgG anti-BSA concentrations (r ₈ =0.77; P <0.0001).

The concentration of long lived IgG anti-BSA antibodies remained in the diabetic patients 3 to 4 months after diagnosis, the concentration of the short lived IgA anti-BSA antibodies was lower (P <0.001, Table 3) . In the 44 patients studied 1 to 2 years after diagnosis, the concentration of all three types of anti- BSA antibodies was lower (P <0.001), reaching normal levels in most patients (IgG: 27 patients, IgA: 43 patients. Table 2) . These results are consistent with the decline in antigenic stimulation by beta cell p69 ¹³¹⁴ .

Studies of Specificity: Additional studies were done using serum from 44 diabetic children and 44 normal children. IgG antibodies to bovine milk casein (Sigma) and j8-lactoblogulin (Sigma) were measured using coated microspheres as described for BSA. The ABBOS peptide (BSA sequence position 152-168) and the homologous region of rat serum albumin (ABRAS peptide) were synthesized with a C-terminal cysteine residue not present in the natural sequence, and the C-terminal cysteine was biotinylated. Solid phase ABBOS and solid phase ABRAS were prepared by binding the biotinylated peptide to streptavidin coupled to carboxylated polystyrene microspheres, as described by Dosch et al. (1988) above. 20μl (0.5% w/v) of ABBOS-or ABRAS-

conjugated microspheres were mixed with 3μl patient or normal serum in a final volume of 0.3 or 3 ml. After incubation at 4°C for 15 minutes, the mixtures were centrifuged and 20μl of the supernatant was used for measurement of residual anti-BSA antibodies.

Anti-ABBOS Antibodies: Measurement of anti-BSA antibodies before (Fig. 2, black bars) and after exposure of serum to solid phase ABBOS peptide (white bars) was done to determine the proporation of anti-BSA antibodies that specifically bound the ABBOS region of BSA.

These studies were done using serum from 44 diabetic patients with high, moderate and realtively low anti-BSA concentrations and an average near the overall mean (Fig. 2, 17 representative samples). The concentration of IgG anti-BSA antibodies decreased by two-thirds (range 30 to 70%) after reaction of serum with ABBOS peptide. Similarly, a large proportion of the IgA and IgM anti-BSA antibodies (23 to 71%) were ABBOS- specific (Table 3), delineating a severe bias for this short sequence that represents less than two percent of the BSA molecule. The amount of anti-BSA antibody removed by incubation of serum with the ABRAS peptide was within normal assay variation (-10%) .

The decrease in anti-BSA antibody concentrations after diagnosis was initiated by the disappearance of ABBOS-specific antibodies (90%) , P <0.0001). By 1 to 2 years after diagnosis only 7 of the 44 patients studied had anti-BSA antibodies with specificity for the ABBOS peptide and 17 of the 44 patients had slightly elevated anti-BSA concentrations. Serum samples from the 17 normal children with the highest anti-BSA concentrations were studied similarly

after incubation of their serum with the solid phase ABBOS peptide. The anti-BSA antibody concentrations were not significantly reduced in the absorbed serum, and only 2 serum samples contained detectable IgG or IgA anti- ABBOS antibodies, respectively. In 300 adult blood donors from Toronto, the range and mean of IgG anti-BSA antibody concentrations were similar to that in the 79 normal Finnish children (Table 3), and IgG anti-ABBOS antibodies were found in three percent of the samples.

Anti-BSA Antibodies and Disease Markers: No relationships were found between the concentration of anti-BSA or anti-ABBOS antibodies and the severity of disease presentation (blood glucose, HbAj, serum C-peptide concentrations) , the duration of symptoms before diagnosis, or the severity of diabetic ketosis or acidosis. The specificity, concentrations and isotype distribution of the antibodies were similar in multiplex and simplex families.

At the time of diagnosis 78% of the patients were islet cell antibody positive, 58% had complement- fixing islet cell-, and 47% had insulin autoantibodies. Niether anti-BSA concentrations, isotype diversity nor specifity were associated with the presence or absence of islet cell- or insulin autoantibodies.

Children heterozygous for HLA-DR3/4 or -Dw3/4 initially had more severe diabetes (higher blood glucose and HbAj and lower serum C-peptide concentrations) than those negative for such haplotype combinations, but the frequencies or concentrations of insulin- and islet cell autoantibodies were similar in both haplotype groups. The concentrations of BSA/ABBOS antibodies were comparable among the diabetic children with or without HLA-DR3/4 or -Dw3/4, as well as in those with -DR3/x, -DR4/x and -Dw4/x.

Solid phase ABBOS was also used to assay serum anti-ABBOS antibodies as described above for use of solid phase BSA (BSA coated microspheres) .

Monoclonal antibodies binding both BSA and

ABBOS (anti-BSA/ABBOS antibodies) were generated from Epstein-Barr virus transformed B lymphocytes of a newly diagnosed diabetic patient (24) . Monoclonal antibodies as well as polyclonal rat anti-ABBOS antisera (19) gave negative reactions for insulin and islet cell autoantibodies. Conversely, addition of up to 1000 and 100 μg of BSA or ABBOS peptide respectively did not alter the results of assays for islet-cell antibodies and insulin autoantibodies in the serum of all 15 diabetic patients tested.

Example 2 - Comparison of PCFIA and ELISA

Patients: Forty Finnish diabetic children (22 males, mean ± SD age 6.2±4.5 years, range 0.9-15.5 years) were randomly selected from the patient population of Example 1 for assay comparison and contained a typical range of elevated levels of BSA antibodies (Fig. 3A) . Samples were drawn at the time of diagnosis of diabetes, and selected sera were sent coded back to the laboratory in Finland to be analysed by EIA. Control subjects comprised 179 age- and sex-matched non-diabetic Finnish children (98 males, mean ± SD age 6.2±3.6 years, range 0.9-15.9 years)

Anti-BSA antibodies in patient and control sera were measured by PCFIA as described in Example 1.

An in-house (positive) standard serum pool from diabetic children was used and a standard curved prepared in every plate (Fig. 3B) . Competition experiments showed

satisfactory specificity for BSA. Free BSA blocked antibody binding in a dose-dependent fashion, whereas neither ovalbumin nor Tween-20 nor both together were able to displace antibody (Fig. 3D) . Results are expressed as kilo fluorescence units (KfU) per microlitre based on instrument gain (5x) , serum dilution and assay volume as derived from the standard containing 12,300 KfU/ml IgG- and 4,200 KfU/ml IgA-anti-BSA antibodies. Due to the linearity of fluorescence emission energy the values were normally distributed. Elevated antibody levels were defined to exceed the mean level in control subjects plus 2 SD.

Enyme Linked Immunosorbent Assay (ELISA)

The serum samples assayed by PCFIA as described above were also assayed by an enzyme linked immunosorbent technique for anti-BSA antibodies, a conventional three- layer solid phase procedure modified from Tainio et al. (25) .

The method employs polystyrene Microstrip wells (Labsystems, Helsinki, Finland) that are processed in an automatic EIA analyser (Auto-EIA II; Labsystems, Helsinki, Finland) , which can process up to three plates (66 samples) per day. The plates were coated with 2 μg/ml (100 μl) BSA (A-4378; Sigma) in 0.1 mol/1 PBS-5 mol/lNaN ₃ (pH 7.4) overnight at room temperature. After washing with PBS-NaN ₃ , the wells were saturated with 1% gelatin-PBS-NaN ₃ for 1 h at 37°C and stored at +4°C until used.

Serum samples were diluted 1:40 in 0.5% gelatin-PBS-NaN ₃ prepared in 0.05% Tween-20. Three replicates of lOOμl of serum dilutions were plated, two in coated wells and one in a non-coated well. After a 60 min incubation at 37°C wells were washed (4x) . Diluted alkaline phosphatase conjugated anti-human IgG- or -IgA

(cat. no. 67806 and 67808; Or-ion Diagnostica, Espoo, Finland) was added and incubated for 60 min followed by four washes as before. After 45 min incubation of substrate [pNPP (p-nitrophenyl phosphate) 2mg/ml in DEA (N-N-diethylaniline)-buffer] the reaction was stopped with 1 mol/1 NaOH. Absorbances (optical density 405nm) in non-coated wells were subtracted from test values.

Intra-assay and inter-assay variations of the assay were 9.3% and 15.8%, respectively. For serum assays, serial dilutions of a BSA-antibody positive standard were run on each plate (Fig. 3C) and the results expressed as percent binding of the standard serum. Values for both IgG- and IgA were skewed despite log- transformation, and thus the limit for positivity was set at the 90th percentile of the values in control subjects. This limit was selected after examining a series of cut¬ off values as giving the highest sensitivity with acceptable specificity. Sera having an absorbance of 3.9% for IgG and 14.2% for IgA of the standard were considered as positive.

Statistical analysis

Statistical analysis was performed using cross- tabulation, Chi-Square statistics and Student\'s unpaired t-test in the case of normally distributed variables. Since the distribution of BSA antibody levels obtained by EIA was skewed despite transformation, the difference between diabetic and control children was evaluated by a Mann-Whitney U-test. Mann-Whitney U-test and Spearman\'s rank correlation test were used to compare antibody levels between I \ and PCFIA. Sensitivity and specificity of the assays was determined, and the results evaluated by cross-tabulation with chi-square statistics. Results are presented as means ± SEM.

The 40 sera from diabetic children contained a range of elevated IgG-anti-BSA ^PCFIA antibodies; however, only 25% were found positive by EIA (Table 4, p <0.0001). Conversely, IgG-anti-BSA ¹ ^ antibodies in control subjects were elevated more frequently than those detected by PCFIA (Table 4, p <0.0002). Elevated IgA-anti-BSA antibodies in diabetic children were found in 50% and 42% (p=NS) , however only 3.3% and 10% of control children were positive in PCFIA and EIA, respectively (Table 4, p <0.01). These results demonstrated that PCFIA but not EIA preferentially detects disease-associated BSA antibodies in children with Type 1 diabetes. In contrast, EIA shows preference for detection of antibodies more prevalent in the general population. Neither procedure detected all BSA antibodies.

BSA antibody levels in diabetic children are shown in Figure 4. There was a significant difference in the levels of both IgG-, and IgA-anti-BSA antibodies between diabetic and control children when determined by PCFIA (p <0.0001 and P <0.001). In contrast, the levels of IgG-anti-BSA^ antibodies were roughly similar in diabetic and control children. IgA-anti-BSA^ antibodies were higher in diabetic children, but the difference (p <0.01) was less prominent than in PCFIA (p <0.001).

These findings suggest a quantitative difference in the subsets of antibodies detected by PCFIA, and this difference distinguishes diabetic and control children.

Individual PCFIA- and EIA values are compared in Figure 5 for diabetic and control children. Shaded areas indicate the levels considered as negative (see above) . The correlation between PCFIA and EIA was very poor (-0.05 s <0.28, 0.09

subjects only one IgA sample (0.3%) was positive in both assays.

BSA-antibodies in control sera bound significantly more frequently in EIA than in PCFIA (IgG: p <0.002 and IgA: p <0.01, Table 4) and only one out the 18 control subjects positive for IgG- or IgA-anti-BSA^ was positive by PCFIA. On the other hand, of the four IgG- and six IgA- positive control sera detected by PCFIA, only one was positive by EIA (Fig. 5) . Therefore, most disease-associated BSA-antibodies were only detected by PCFIA. The sensitivity of PCFIA in detecting disease- associated IgG anti-BSA antibodies was excellent when compared to EIA (100% vs 25%, Table 5; p <0.0001) and the disease-specificity was higher for PCFIA (Table 5; p <0.02). For the IgA isotype assay results were comparable with respect to disease-sensitivity, but in EIA this was at the cost of assay specificity (p <0.05). These findings suggest that PCFIA and EIA preferentially detect different subsets of BSA antibodies with no (IgG) or some overlap (IgA) . Antibodies detected in EIA are not associated with Type 1 diabetes but are found commonly in the general population.

TAB LE 4

Table . Frequency of elevated anti-BSA antibodies in diabetic (n=40) and control children (n= 179) as defined by particle concentration fluoroimmunoassay (PCFIA) and by enzyme immuno¬ assay (EIA).

0)

C Isotype Patients Positive Positive Control Subjects

CD ω n .%) n (*

H PCFIA EIA P PCFIA EIA

H t

C

^ m IgG 40» (100%) 10* (25%) <0.0001 4 (2.2%) 18 (10%) <0.002

CO

I IgA 20* (50%) 17* (42%) NS 6 (3.3%) 18 (10%) <0.01 m m

-4 For difference between antibody positive patients and control subjects using PCFIA and

TABLE 5

Table Sensitivity and specificity of PCFIA and EIA in detecting anti-BSA antibodies.

SUBSTITUTE SHEET

In order to compare the PCFIA assay procedure to the more commonly available EIA, we analysed a large number of samples using both techniques. The comparison revealed unequivocal differences with diabetes-associated anti-BSA molecules detected almost exclusively by PCFIA. Both, levels and frequency of positive responses among diabetic children were significantly higher in PCFIA than EIA. The wide scatter of antibody levels in EIA for both patients or control subjects caused major overlap between the groups and made differences statistically insignificant. A large proportion of samples were below the detection limit of EIA, causing considerable skewness for measurement of both isotypes examined.

PCFIA detected BSA antibodies in children with diabetes, whereas only a few non-diabetic children were positive. In contrast, only a small proportion of the antibodies were disease-associated in EIA, and levels were elevated more often in non-diabetic subjects. With only one out the 18 controls positive for IgG- or IgA- anti-BSA ^EtA elevated by PCFIA as well, the dichotomy was clear between antibody subsets detected in either procedure. Interestingly, the single elevated IgG and IgA values detected in both procedures derived from the same serum sample, suggesting that the host determines the choice of antibody species utilized in the common immune response to dietary BSA.

The BSA molecule consists of 608 amino acids and there are several areas where the sequence differs from human serum albumin. One of those is the described ABBOS (pre-BSA position 153-169 (19)).

As seen in example 1, most of the diabetes- associated antibodies detected by PCFIA in diabetic children are directed against this epitope, whereas in non-diabetic control subjects the major epitopes are

different, with less than 3% of donors able to recognize ABBOS. Since the ABBOS epitope is immunologically cross- reactive with a (beta-cell) autoantigen, p69 [1,3], the poor immunogenicity of this epitope in the general population is not surprising and clearly identifies the diabetic population. We have tentatively linked this principal difference to efficient antigen presentation of ABBOS by diabetes-associated MHC class II molecules coupled with a delay in oral (or mucosal) tolerance development in diabetic subjects (14,13): our focus on the latter was triggered by the report that the single highest marker of diabetes risk (OQβ non-ASP ⁵⁷ ) also marks susceptibility for IgA deficiency, a regulatory abnormality of mucosal immunity (26) .

An antigen such as BSA has several epitopes which can induce a wide spectrum of high and low affinity antibodies. Our results are very reminiscent of the observation that insulin-autoantibodies (IAA) detected by EIA are poorly disease-associated (27) and have a low predictive value compared to fluid-phase radiobinding assays (RIA) (28-30) . EIA has been characterized by low sensitivity and an unacceptably high rate of false positives (27) , similar to the results obtained in this study. EIA detects mainly low affinity IAA that have high binding capacity, whereas RIA detects high affinity, low binding capacity and strongly disease-associated subset(s) of IAA (30) . The same could apply to a fluid- phase assay such as PCFIA, in which accessibility to the epitopes may be different from EIA. The same epitopes may not be available in PCFIA and EIA due to different binding procedures. Disease-associated epitopes may not be accessible if bound to EIA plate (31) . On the other hand, an excess of adhesive antigens on EIA-plate surface aay . as binding of non-disease-associated, low affinity anti_._>dies with high binding capacity as reported for IAA in healthy blood donors (32) .

The striking lack of correlation between the two assay systems is less suggestive of gradual differences in average antibody affinity, but indicates absolute distinctions in the quality of antibodies detected. Maturation of an antigen drive (hyper-) immune response produces an antibody repertoire that is not only of high affinity but also favors immunoglobulins with fast binding kinetics (33) , i.e. antibodies characterized by high on:off antigen binding rates and release required, for example, for rapid opsonization of pathogens and re-utilization of antibody. We speculate that the combination of large-surface area of antigen- conjugated microspheres, consequent ease of antigen accessibility and the fast dynamics of PCFIA (1 mi binding periods) all contribute to the preferential detection of antibodies with a high on:off binding rate. The observations presented here emphasize the importance of clinical validation for serological assay procedures which rarely cover all possible immunoglobulin repertoires able to associate with a given antigen.

Example 3 Detection of pre-clinical IDDM by PCFIA

Patients and Cases: The samples in this study were obtained as a subset of the blood samples employed in Example 1 which included over 90% of all families with a newly diagnosed child with IDDM ("INDEX CASE") over a period of several years (the study is still ongoing) . Blood samples were taken from siblings of index ceases at the time of diagnosis for the latter and every six-twelve months later. Over 700 families were enrolled and 19 siblings of index cases turned diabetic so far. Our samples represent a subset of 11 of these converters and one hundred siblings which appear healthy (at most 1-2 of these are expected to convert to diabetes) . Of all these children there were 1-6 samples available taken at the above interval.

Serum IgG anti-BSA antibody levels were measured as described in Example 1.

The results are set out in Tables 6 & 7.

Table 6 shows IgG anti-BSA antibody levels present in the very first sample taken from each subject (i.e. when the \'index case\' was diagnosed). Diabetes developed in these children 1-5 years later. Table 7 shows IgG anti-BSA antibody levels in a random subset of 100 healthy siblings to the index cases.

O 94/02507

30

TABLE 6

^Δ Elevated levels are defined as levels above the Mean +2 Standard Deviations (SD) found in normal controls. The subset of control sibling samples shown here (last line) is not significantly different from levels observed in matched, random population controls (i.e. samples from healthy children).

♦Random subset of 1QQ healthy siblings to index cases

TABLE 7

TABLE 7 CONT\'D

Example 4 Detection of BSA-sensitised IDDM-associated T- lymphocytes

Patients: Venous blood samples were obtained with informed consent from patients with recent onset IDDM and from healthy controls. T lymphocytes were prepared from the blood samples and T cell proliferation in response to BSA and to various synthetic peptide fragments of -BSA was assayed as described in Dosch et al. (22), but wi€h substitution of serum free media.

Synthetic peptides were generated on a Pharmacia peptide synthesiser (Pharmacia LTD, Montreal, Quebec) following the manufacturers\' recommendations and purified by conventional HPLC techniques.

The following synthetic peptides were developed:

ABBOS: ac-Phe-Lys-Ala-Asp-Glu-Lys-Lys-Phe-Trp-Gly-Lys-

Tyr-Leu-Tyr-Glu-Ile-Ala-Arg-Arg—Cys-0-NH ₂ fABBOS (alias CS2185) : FKADEKKFWGKYLYEIARR)

ABRAS: ac-Gln-Glu-Asn-Pro-Thr-Ser-Phe-Leu-Cys-His-Tyr- Leu-His-Glu-Val-Ala-Lys-Lys-NHj

(ABRAS (no alias ) : OENPTSFLCHYLHEVARRl

CS2270: ac-Tyr-Ala-Asn-Lys-Tyr-Gly-Val-NH ₂ (YANKYGV)

CS2268: ao-Lys-Phe-Trp-Gly-Lys-Tyr-NH ₂ (KFWGKY)

CS2267: ac-Glu-Phe-Lys-Ala-Asp-Glu-Lys-Lys-NH ₂ (EFKADEKIO

CS2240: ac-Ile-Glu-Thr-Met-Arg-Glu-Lys-Val-Leu-Thr—Cys-

0-NH ₂ (IETMREKVLT)

Briefly, heparinized blood is diluted 1:1 with serum-free HL/1 medium and mononuclear cells are purified on Ficoll-Hypaque gradients, washed in HL/1, counted and 2xl0 ⁵ are added to 96 well microculture plates in a volume of 200 μl HL/1. Control cultures receive either no or a supplement of the pan-T cell mitogen Phytohemagglutinin (lμg) . BSA (0.01-10μg) or other test proteins, or 0.1- lμg of a synthetic peptide are added to triplicate test cultures. Cultures are incubated for 8-10 days at 37°C in a humidified atmosphere of 5% C02 in air until pulsed for 6 hrs with lμCi ³ HTdR. Cells are then harvested with an automated harvester and incorporated thymidine is measured by scintillation counting. Alternative measures of T cell activation such as CCFA or measurements of IL2 production etc. are equally suited. All data presented here employed the above, standard TdR incorporation procedure.

The results from IDDM patients are shown in Table 8 and those from normal controls in Table 9.

All data are expressed as proliferative responses relative to that induced by PHA [ (cpm experimental/cpm PHA)*100]. A positive response is defined as mean response in unstimulated cells plus 2 SD. All patients but no controls showed a positive response to at least BSA or ABBOS, usually both, and, where tested, to CS 2267, the minimum T cell stimulatory peptide under our conditions. Occasional small responses were seen to CS2268 but not to CS2270 or CS2240.

TABLE 8

SUBSTITUTE SHEET

:ABLE 8 CONT\'D

SUBSTITUTE SHEET

33 ABLE 9

TABLE 9 CONT\'D

IDDM patients were found to have sensitized T lymphocytes which specifically recognize and proliferate in response to peptide CS 2267 while normal controls lacked such sensitized T lymphocytes. Subjects with pre- clinical IDDM as assessed by currently available indicators showed the same response as IDDM patients.

Having established the predictive power of anti-ABBOS and anti-BSA immunity, the present inventors have also developed an in vitro system to detect T lymphocytes which initiate and sustain this immunity until final β cell demise.

References

I. Elliott, R.B., Reddy S.N., Bibby, N.J., Kida, K. (1988), Diabetologia 36: 62-64. 2. Borch-Johnsen, K. , Joner, G. , Mandrup-Poulsen, T. et al., (1984), Lancet II: 353-60.

3. Scott, F.W., (1990), Am. J. Clin. Nutr. 51(3): 489- 91.

4. Virtanen, S. , Rasanen, L. Aro, A. et al. (1991), Diabetes Care 14(5): 415-417.

5. Scott, F. ., Cloutier, H.I]. , Souligny, J. , Riley, W.J., Hoorfar, J. , Brogen, CH. , (1989), Diet and antibody production in the diabetes-prone BB rat. In: I icins, R. , Zimmet, P., Chisholm, D. (eds) Diabetes 1988. Proceedings of the 13th

International Diabetes Federation congress, Elsevier Science Publishers, Amsterdam, pp. 763-767.

6. Elliott, R.B., Martin, J.M. , (1984), Diabetologia 26(297): 297-299. 7. Dane an, D., Fishman, L. , Clarson, C. , Martin J.M. , (1987), Diabetes Res. 5(93): 93-97.

8. Scott, F.W. , Marliss, E.S., (1991), Can. J. Physiol Pharmacol 69(3): 311-319.

9. Beppu, H. , Winter, W.E., Atkinson, M.A. , Maclaren, N.K., Fujita, K., Takahashi, H. , (1987), Diabetes

Res 6: 67-69.

10. Yokota, A., Ya aguchi, Y., Ueda, Y. et al. (1990) Diabetes Res. Clin. Prac. 9: 211-217.

II. Savilahti, E., Akerblo , H.K. , Tainio, V-M. , Koskimies, S., (1988), Diabetes Res. 7: 137-140.

12. Pocecco, M. Nicoloso, F. , Tonini, G. , Presani, G. , Marinoni, S. (1991), Hor . Res., p. 67.

13. Martin, J.M. , Trink, B., Daneman, D. , Dosch, H-M. , Robinson, B.H. , (1991), Ann. Med. 23: 447-452. 14. Dosch, H-M., Karjalainen, J. , Morkowsky, J., Martin, J.M. , Robinson, B.H. , (1992), Nutritional triggers of IDDM. In: Levy-Marshall, C, Czernichow, P. (eds)

Pediatric and Adolescent Endocrinology. Karger: Basel, p. 202-217. 15. Pilcher, C.C., Dickens, K. , Elliott, R.B. , (1991), Diabetes Res. Clin. Pract. 14 (Suppl. 1) : 82A. 16. Karjalainen, J.K. , Diabetes 1990; 39: 1144-50.

17. Karjalainen, J. , Knip, M. , Mustonen, A., Akerblom, H.K., Diabetes, 1988; 31: 129-133.

18. Karjalainen, J. , Knip, M. , Mustonen, A., Ilonen, J. , Akerblom, H.K., Diabetes, 1986; 35: 620-2.. 19. Glerum, M. , Robinson, B.H., Martin, J.M. , (1989), Diabetes Res. 10: 103-107.

21. Dosch, G-M, Lam, P., Guerin, D. , Leeder, J.S., (1988) , Characteristics of particle concentration fluorescence immunoassay (PCFIA) : Novel alternatives to ELISA and RIA, In: Gembala P-L (ed) Proceedings of the 1987 Pandex Symposium on Particle Concentration Fluorescence Immunoassay, Baxter Healthcare Corporation, Chicago, pp. 5-22.

22. Dosch, H-M., Lam, P., Hui, M.F., Hibi, T. , Cheung, R.K., (1990), Int. Immunol 2 (Sept. 1990): 833-848.

23. Cheung, R.K. , Dosch, H-M., (1991), J. Boil. Chem. 266(14): 8667-8670.

24. MacKenzie, T., Dosch, H-M., J. Exp. Med. 1989; 169: 407-30. 25. Tainio, V.M. , Savilahti, E. , Arjomaa, P.,

Salmenpera, L. , Perheentupa, J., Siimes, M. , (1988),

Acta. Paediatr. Scand. 77: 807-811. 26. Olerup, O. , Smith, C.I.E., Ha arstrδm, L. (1990),

Nature 327: 289-290. 27. Caudwell, J. , Stojanovski, C. , Colma, P. (1991),

Diabetes Res. Clin. Pract. 14 (Suppl. 1) : 89A. 28. Kuglin, B., Kolb, H., Greenbaum, C. , Maclaren, N.K.,

Lernmark, A., Palmer, J.P., (1990), Diabetologia 33:

638-639. 29. Levy-Marchal, C. , Bridel, M.P., Sodoyez-Goffaux, F. et al., (1991), Diabetes Care 14 (Jan): 61-63.

30. Greenbaum, G.J., Palmer, J.P., Kuglin, B. , Kolb, H. , (1992), J. Clin. Endocrinol Metab 74: in press.

31. Sodoyez-Goffaux, F. , Koch, M. , Dozio, N. , Brandenburg, D., Sodoyez, J-C, (1988), Diabetologia 31: 694-702.

32. Sodoyez, J-C, Sodoyez-Goffaux, F., Koch, M. , Bouillenne, C, Francois-Gerard, C, Bosi, E. , (1990), Diabetologia 33: 719-725.

33. Foote, J. , Milstein, C, (1991), Nature 352: 530- 532.