(Received for publication, December 5, 1996, and in revised form, February 19, 1997)
From the Department of Medical Biochemistry, We describe the application of a stable isotope
dilution assay (IDA) to determine precise insulin, C-peptide, and
proinsulin levels in blood by extraction from serum and quantitation by
mass spectrometry using analogues of each target protein labeled with stable isotopes. Insulin and C-peptide levels were also determined by
immunoassay, which gave consistently higher results than by IDA, the
relative difference being larger at low concentrations. Insulin,
C-peptide, and proinsulin levels were all shown by IDA to be higher in
type II diabetics than in non-diabetics, with mean values rising from
22 (± 2) to 92 (± 8), 335 (± 11) to 821 (± 24), and 6 (± 1) to 37 (± 3) pM, respectively. Interestingly, the ratio
between IDA and immunoassay values for insulin levels increased from
1.3 in non-diabetics to 1.7 in type II diabetics. The ratio between
proinsulin and insulin levels by IDA increased from 0.24 in
non-diabetics to 0.36 in type II diabetics, whereas the ratio between
C-peptide and insulin levels by IDA decreased from 17.6 to 10.7. This
disproportionate change in protein levels between different types of
individuals has implications for the metabolism of insulin in the
diabetics studied (type II) and suggests that C-peptide levels are not
always a reliable guide as to pancreatic insulin secretion. In
addition, levels of the 33-residue C-peptide (partially trimmed form)
were shown to be less than 10% that of the fully trimmed 31-residue
C-peptide levels, and we tested IDA in a clinical context by two
post-pancreatic graft studies. IDA was shown to give direct, positive
identification of the target protein with unrivaled accuracy, avoiding
many of the problems associated with present methodology for protein
determination.
At present, protein levels in humans are measured by following
radiolabeled material or by the use of immunological reagents. Radioiodination is the most common procedure of protein labeling and
has been of great value. However, it has been recognized that chemical
modification of a protein in this way can affect its biological and
physicochemical behavior. Tissue dehalogenases, which are known to
remove iodine from iodotyrosine, may invalidate concentration
measurements, particularly of fragments of the target molecule (1, 2).
Tritiated proteins have been used since they escape these objections
(3), but the ethical issue of injecting radioactive proteins of any
kind into human subjects remains under severe scrutiny.
The alternative method involving the use of immunological reagents to
estimate blood protein levels remains the best current approach.
However, antibodies used in these measurements often cross-react with
precursors of the target protein or with its smaller, degraded
fragments, creating uncertainty in results (2, 4-6). Although modern
forms of radioimmunoassay, microparticle enzyme immunoassay
(MEIA),1 and immunoradiometric assay (7-9)
reduce many of these cross-reactivities, the methodology still remains
an indirect measure of the target molecule, where endogenous and
administered exogenous protein cannot be distinguished. Also,
pathological samples may contain endogenous antibodies that can
interfere with conventional assays and invalidate the results
obtained.
Such difficulties, as shown by significant discrepancies between
immunoassays performed at different centers on the same samples (10),
have led to a feeling that measurements of insulin levels, for example,
have been systematically overestimated by radioimmunoassay in type II
(non-insulin-dependent) diabetic individuals, resulting in
insufficient attention being paid to insulin deficiency as opposed to
insulin resistance (5). Indeed, many diseases, including type II
diabetes and tumors such as insulinomas or certain other carcinomas,
lead to a disproportional increase in the concentration of proteins
such as proinsulin and its conversion intermediates that are
immunologically cross-reactive with insulin (6, 11-13). The uncertain
clinical interpretation of results involving such important diseases
remains a serious problem.
A definitive measure of blood protein concentration would be most
welcome in calibrating existing procedures and investigating suspect
results. To be able to fully understand the fate and distribution of
proteins in vivo, a technique is required that can
accurately determine each of the components directly involved, allowing
a distinction between endogenous and exogenous protein while avoiding the injection of radioactive material. To this aim we describe an
alternative technique, referred to as isotope dilution assay (IDA),
that allows precise quantitation of specific proteins at physiological
concentrations in blood.
In this procedure, target proteins are extracted from a 1-ml serum
sample by solid-phase or immunoaffinity techniques and purified by
HPLC. Their initial concentration is then determined by mass
spectrometry from the ratio between the spectral peak of each target
protein and that of an internal standard that has been added to the
serum in known amounts. A carrier protein is also added in excess
concentration, which allows easier detection of the three proteins by
HPLC and helps to avoid protein loss by adsorption. The internal
standards and carrier proteins are stable isotope-labeled analogues of
the target proteins that exhibit slightly different mass.
IDA incorporates a method known as stable isotope dilution mass
spectrometry that has proved successful for the measurement of small
molecules (14-17). However, its use for macromolecules such as
proteins is less developed due to difficulty in the synthesis of the
necessary internal standard and carrier analogues. Having previously
demonstrated the feasibility of our approach by determination of
insulin levels in serum (18), we now report the development of an
optimized version of this technique to measure in humans the protein
pair proinsulin and insulin together with the activation-related connecting peptide (C-peptide) (13, 19-22).
The precursor of insulin, proinsulin, is produced by the beta cells of
the pancreatic islets of Langerhans. Most of the proinsulin is then
enzymatically cleaved to yield insulin and the C-peptide via molecules
known as conversion intermediates (split proinsulins), but some is
secreted from the pancreas intact. Insulin is enzymatically digested,
largely in the liver, possibly by the insulin-degrading enzyme. The
relative levels of proinsulin, insulin, C-peptide, and the four
conversion intermediates (32-33, 65-66, des-31-32, and des-64-65
split proinsulins) are known to vary significantly between different
types of individuals (4, 13, 23-26). However, the precise difference
between levels of these proteins is difficult to evaluate since present
methodology cannot yet distinguish between all of them. Increased
knowledge about the secretion, mode of action, and distribution of
insulin-related proteins in the body is required to understand
important diseases such as diabetes. Using IDA, we investigated
proinsulin, insulin, and C-peptide levels in non-diabetic non-obese,
non-diabetic obese, and type II diabetic individuals and compared our
results with those obtained by standard immunoassay.
Materials
Carboxypeptidase B (from porcine pancreas) was obtained from
Boehringer Mannheim. Trypsin (TPCK-treated, from bovine pancreas), antiserum against human insulin (developed in guinea pig), and bovine
serum albumin (essentially fatty acid and globulin free) were obtained
from Sigma. All other reagents (except the antiserum against human
C-peptide, see "Acknowledgments") were of analytical grade and were
purchased from either Fluka Chemie AG or Sigma.
Preparation and Purification of Protein Analogues
The proinsulin gene, in a construction containing a
polyhistidine tag (27), was cloned and expressed in Escherichia
coli by a modified (see "Acknowledgments") standard procedures
(28) at a yield of 5-10 mg/liter of culture. After protein isolation using a 10-ml nickel-Sepharose fast flow column (Pharmacia Biotech Inc.), the polyhistidine tag was removed with cyanogen bromide (1.5 mg/mg of protein at pH 1), and proinsulin was purified by reverse-phase
HPLC (using a Macherey Nagel 300-Å C8 column (250 × 10 mm)) with
a linear gradient of acetonitrile (22-36% in 30 min) in 0.1%
trifluoroacetic acid in water. Proinsulin was labeled biosynthetically
with 15N by expression of the gene in a culture medium
containing 15NH4Cl (28, 29), where the
incorporation of 15N was calculated to be 98.8% by mass
spectrometry. Furthermore, proinsulin was double-labeled with both
15N and 2H by expression in
15NH4Cl and 50% 2H2O
medium, where the final incorporation of 2H after protein
purification was calculated to be 24.3% (assuming 98.8%
15N incorporation). Each proinsulin analogue was cleaved by
proteolytic techniques, using trypsin (2.5 µg/mg of protein) and
carboxypeptidase B (5 µg/mg of protein) at pH 8 to yield
15N- and 15N2H-labeled insulin and
C-peptide (31- and 33-residue forms).
[2H8]PheB1-[2H8]ValB2
insulin ([2H16]insulin) was prepared as
described previously (30).
[2H2]Gly7 C-peptides
([2H14]C-peptides) (31- and 33-residue forms)
were prepared by total peptide synthesis using standard Fmoc
(N-(9-fluorenyl)methoxycarbonyl) chemistry (Applied
Biosystems 430A peptide synthesizer). Subsequent peptide purification
was performed by reverse-phase HPLC using similar conditions as
described above.
The concentration of proinsulin and insulin solutions was determined
from the absorbance at 280 nm (31), and the concentration of C-peptide
solutions was determined by amino acid analysis, HPLC, and electrospray
ionization mass spectrometry (ESI-MS) against a standard. Protein
aliquots were prepared at a concentration of 25 µM,
freeze-dried, and stored at Extraction of Proteins from Serum
After dilution with 5 ml of 0.1%
trifluoroacetic acid in water, the serum sample (1 ml) was filtered
through a 0.45-µm sterile filter (Millipore Corp.), and 150 IU of
heparin was then added (to bind PF4, a platelet factor in
serum that has a HPLC retention time similar to insulin). A known
amount of each internal standard (at a concentration similar to that
expected for the target protein) and each carrier (at excess
concentration; 100 nM [15N]insulin, 200 nM [15N2H]C-peptide, and 100 nM [15N2H]proinsulin) were added
to the serum. A Sep-Pak cartridge (Waters C18) was washed with pure
acetonitrile and then equilibrated in 0.1% trifluoroacetic acid in
water. The serum was passed back and forth through the cartridge three
times at approximately 1 ml/min using polypropylene syringes fixed at
either end. Two solutions (4 ml) consisting of 20% acetonitrile, 0.1%
trifluoroacetic acid in water and 80% acetonitrile, 20%
dichloromethane were then washed through the cartridge to remove
unwanted proteins (32). The target insulin proteins were co-eluted from
the cartridge with 50% acetonitrile, 0.1% trifluoroacetic acid in
water (2 ml), and the solution was freeze-dried.
After dilution with 5 ml of 0.1 M sodium phosphate buffer, pH 8, the serum sample (1 ml)
was filtered through a 0.45-µm sterile filter (Millipore), and 150 IU
of heparin and a known amount of each internal standard and each
carrier were then added (as for solid phase). The serum was passed
twice (under gravity) through an affinity column pre-equilibrated in
the phosphate buffer, and the column was then washed with the buffer (6 ml). The column consisted of an Aminolink-agarose gel (3 ml,
Pierce) to which both insulin antibodies (guinea pig) and C-peptide
antibodies (mouse IgG, known to also bind proinsulin) were coupled at
excess concentration (1 mg/ml). A Sep-Pak cartridge (Waters C18 light) was washed with pure acetonitrile, equilibrated in 0.1%
trifluoroacetic acid in water, and fixed onto the outlet of the
affinity column. A solution of 2 M acetic acid (6 ml) was
passed down the column to elute the target proteins into the Sep-Pak
cartridge. The cartridge was then removed and washed with 20%
acetonitrile, 0.1% trifluoroacetic acid in water (4 ml). The target
insulin proteins were eluted from the cartridge with 50% acetonitrile,
0.1% trifluoroacetic acid in water (2 ml), and the solution was
freeze-dried.
Purification of Proteins by Microbore HPLC
Following extraction by either solid phase or immunoaffinity,
the samples were redissolved in 0.1% trifluoroacetic acid in water
(120 µl) and loaded onto a microbore reverse-phase HPLC system
(Applied Biosystems 140B, Nucleosil 300-Å C18 column (1 × 150 mm)). Each target protein extracted from the serum was then isolated
and purified, together with its internal standard and carrier proteins,
by elution with a linear gradient of acetonitrile (22-36% in 60 min
at a flow of 40 µl/min) in 0.1% trifluoroacetic acid in water. The
excess concentration of the carrier allows easy detection and
purification of the three proteins (the target protein, its internal
standard, and its carrier) together as a single peak by chromatography.
The aliquots collected were freeze-dried.
Quantitation of Proteins by Mass Spectrometry
Aliquots containing each target protein, together with its
respective internal standard and carrier proteins, were resuspended in
10 µl of ESI-MS solvent (50% methanol, 49% water, 1% acetic acid).
The samples were then analyzed by ESI-MS carried out in positive
ionization mode using either: (a) VG-Trio 2000 single quadrupole instrument (Fisons Instruments, Inc.) at a flow of 2 µl/min and a scan cycle of 10 s or (b) VG-Platform II
single quadrupole instrument (Micromass) at a flow of 10 µl/min and a scan cycle of 1 s. The spectrum was scanned to cover the most intense signals of the target protein and its internal standard, whereas the large carrier signal was designed to be out of this scan
range. By analysis of the ratio between the peak height of the target
protein and the peak height of the internal standard (added at known
concentration), the exact concentration of the target protein in the
original serum sample was determined. Any loss of product during the
extraction and purification procedure would not affect the ratio.
Automated Immunoassay
Serum samples obtained from the University Hospital, Geneva,
were initially analyzed by automated immunoassay. Insulin concentration was determined with a MEIA, IMx (Abbott Laboratories, Toyko, Japan), which uses mouse monoclonal antibodies against human insulin. C-peptide
concentration was determined with a chemiluminescent enzyme immunoassay
(CEI), Immulite (Diagnostic Product Corp., Los Angeles, CA), which uses
rabbit polyclonal antibodies against C-peptide. The proinsulin
concentration of these samples was not determined by immunoassay.
Verification of IDA Methodology
Fresh serum samples taken from various types of
individuals were obtained from the University Hospital, Geneva. The
concentration of insulin (66 samples) or C-peptide (40 different
samples) was measured by immunoassay. These samples represent most of
the range of possible blood concentration levels: 10-1638
pM insulin and 55-6820 pM C-peptide. To
compare results obtained by the two techniques, the insulin or
C-peptide concentration in these samples was also measured by IDA. The
amount of [2H16]insulin (the internal
standard for insulin) added to each sample was varied from 10 to 1500 fmol, depending on the insulin concentration as determined by MEIA.
Similarly, the amount of [2H14]C-peptide (the
internal standard for C-peptide) was varied from 100 fmol to 6 pmol,
depending on the C-peptide concentration as determined by CEI.
The concentration of insulin, in
four identical 1-ml aliquots taken from a single fresh serum sample,
was determined by IDA to test the variability of results. This
procedure was repeated for three different serum samples at insulin
concentrations of 50, 203, and 640 pM as determined by
MEIA. A similar procedure was carried out to measure by IDA the
concentration of C-peptide in serum samples determined by CEI to be
210, 560, and 1080 pM. Proinsulin concentration was
similarly measured by IDA of serum samples at an MEIA insulin
concentration of 100, 570, and 890 pM.
[2H16]Insulin and
[2H14]C-peptide were added at a concentration
similar to that of the target protein (as determined by immunoassay),
and [15N]proinsulin (the internal standard for
proinsulin) was added at one-tenth the insulin concentration.
Measurement of the 33-Residue C-peptide Levels
The 33-residue form of the C-peptide comprises an additional two
amino acids, lysine and arginine, which remain at its C terminus after
incomplete cleavage from proinsulin. This form is thought to be present
in serum at a far lower concentration than the fully trimmed C-peptide
consisting of 31 residues.
The C-peptide concentration of 10 serum samples was determined by CEI.
The 33-residue C-peptide and the 31-residue C-peptide were then
individually quantitated in these samples by IDA. The amounts of
internal standards added were 100 fmol of
[2H14]C-peptide 33 and 150-4500 fmol of
[2H14]C-peptide 31.
Measurement of Basal Levels in Non-diabetic and Type II Diabetic
Individuals
Following a 12-h fast, fresh serum samples were obtained from
three types of persons: non-diabetic non-obese (25 samples), non-diabetic obese (10 samples), and type II diabetic (20 samples). A
heterogeneous population was examined from both sexes, aged 20-60
years, where obese individuals were defined as those with a body mass
index above 35 kg/m2. The precise basal level of insulin,
C-peptide, and proinsulin in each of these samples was investigated by
IDA. The concentration of insulin (in all samples) and C-peptide (only
in samples from diabetic individuals) was also measured by immunoassay.
The amounts of internal standards added to each sample were 10-300
fmol of [2H16]insulin, 200-2000 fmol of
[2H14]C-peptide, and 10-100 fmol of
[15N]proinsulin.
Two Post-pancreatic Graft Studies
This patient underwent a subtotal
pancreatectomy with autotransplantation of the islets of Langerhans.
Following the operation, insulin and C-peptide basal levels were
measured by immunoassay from blood samples (10 ml) taken every 3 h
for 60 h. These 20 samples were further analyzed by IDA to
determine insulin, C-peptide, and proinsulin levels during this period.
After coagulation in the absence of anticoagulant, serum was separated
from blood samples by centrifugation. The amounts of internal standards
added to each sample were 10-30 fmol of
[2H16]insulin, 500 fmol of
[2H14]C-peptide, and 50 fmol of
[15N]proinsulin.
Following a similar operation
to that described for patient 1, pancreatic reaction was monitored by
an oral glucose tolerance test, where 70 g of glucose in 400 ml of
water was taken orally after a 12-h fast. Insulin and C-peptide levels
were measured by immunoassay from blood samples (10 ml) taken every 10 min for 50 min. For comparison, insulin, C-peptide, and proinsulin
levels were determined in these samples by IDA. The same amounts of
internal standards were added to each sample as in the study
with patient 1.
Preparation of Standards
Each protein prepared comprises the human sequence and was
characterized by amino acid analysis and ESI-MS (Table
I). Importantly, all of these analogues prepared, both
those to be used as internal standards and those that act as carrier
proteins, had exactly the same retention time by HPLC as their native
counterparts and differ in mass due only to isotopic labeling. Further,
the carrier proteins, present in excess concentration, exhibited no
spectral peaks by ESI-MS that might have obscured the measurement of
their respective target protein or internal standard peaks.
Table I.
Mass data for each analogue of insulin, C-peptide, and proinsulin
prepared
ABSTRACT
INTRODUCTION
EXPERIMENTAL PROCEDURES
RESULTS
DISCUSSION
FOOTNOTES
ACKNOWLEDGEMENTS
REFERENCES
20 °C. To help avoid protein loss, any
further protein dilution was carried out in siliconized glass tubes
with a solution of bovine serum albumin (1 mg/ml). Our investigations
on the C-peptide refer solely to the 31-residue form, unless the
33-residue form is specifically mentioned.
Protein
Molecular
massa
Mass/charge of main ionb
Da
m/z
C-peptide 31
Native
3020
1007.7 (3+)
2H14c
3034
1012.3
15N2Hd
3111
1038.0
C-peptide 33
Native
3304
1102.3 (3+)
2H14c
3318
1107.0
15N2Hd
3408
1137.0
Insulin
Native
5807
1162.4 (5+)
2H16c
5823
1165.6
15Nd
5872
1175.4
Proinsulin
Native
9388
1174.5 (8+)
15Nc
9501
1188.6
15N2Hd
9657
1208.1
a
Standard error in each measurement is ± 1.0 Da.
b
From the ion envelope formed by ESI-MS of each protein, the
ion with the largest intensity was used for analysis. Standard error in
each measurement is ± 0.4 m/z units.
c
Analogue used as internal standard.
d
Analogue used as carrier.
Extraction and Purification of Proteins from Serum
Insulin, proinsulin, and the C-peptide were efficiently isolated
and purified from 1 ml of serum by both solid-phase and immunoaffinity extraction methodology, with approximately 80% recovery of the pure
target protein. Both extraction procedures eliminated most of the
unwanted substances in the serum (which would have overloaded the mass
spectrometer) before further purification by microbore HPLC. This
system keeps the working volumes small to avoid protein loss and is
relatively efficient even at low protein concentration. Immunoaffinity
(Fig. 1) allowed a more specific extraction of the
target proteins than solid-phase methodology, which also extracted proteins of similar mass to the C-peptide (such as platelet factors) that elute at an HPLC retention time between that of the C-peptide and
insulin.
Following extraction and purification, an excellent signal-to-noise
ratio was obtained by ESI-MS of each target protein at its
physiological concentration in the presence of its internal standard,
demonstrating the high sensitivity of this method (Fig. 2). By extraction of a series of samples containing
either negligible amounts of each target protein or no internal
standards, we showed that there was no contaminating signal by ESI-MS
from endogenous impurities in the serum (remaining after extraction and
HPLC purification) that would significantly obscure measurements of the
target protein or its internal standard (data not shown).
Verification of IDA Methodology and Comparison to Immunoassay
Serum samples received from the University Hospital, Geneva, were
stored at 20 °C for up to 3 months before analysis by IDA. Trial
analyses by IDA before and after this period were indistinguishable and
showed that there was no significant change in insulin, C-peptide, or
proinsulin concentration.
Three sets of four repeat extractions of either insulin, C-peptide, or proinsulin from the same serum sample demonstrated that the variability of results by IDA is below 10% at all concentrations measured. However, we noted that this variability was greater at lower protein concentrations and estimate that the lowest detectable limit within this error margin is at 1 pM. Protein concentration by immunoassay was calibrated using a series of standards supplied by the manufacturer. Analysis by IDA showed that the insulin or C-peptide concentration of each of these standards was in agreement with the concentration of our internal standards, [2H16]insulin and [2H14]C-peptide, to within 10% (data not shown).
The insulin and C-peptide concentration of trial serum samples analyzed
by IDA ranged from 3 to 1397 pM (insulin) and from 30 to
5786 pM (C-peptide). Comparison of these results with those by immunoassay shows that measurements are higher by immunoassay for
70% (insulin) and 80% (C-peptide) of the samples (Figs.
3A and 4A).
Although the absolute difference between immunoassay and IDA values is
larger at higher concentrations, the relative difference is larger at
lower concentrations (Figs. 3A and 4A, insets). This may partly be due to larger errors in both
techniques at lower concentrations. There is a linear relationship
between immunoassay and IDA values with slope 0.88 ± 0.03 (insulin) and 0.82 ± 0.03 (C-peptide) (Figs. 3B and
4B). Using this relationship, theoretical values by IDA were
calculated from immunoassay values. A plot of the relative residual
between actual and theoretical IDA values is approximately symmetrical
about zero at all concentrations, indicating that the bias is not
concentration-dependent (Figs. 3B and
4B, insets).
Measurement of the 33-Residue C-peptide Levels
The C-peptide concentration of the 10 samples analyzed ranged from 150 to 4850 pM as determined by CEI. This does not distinguish between the 31- and 33-residue forms of the C-peptide that were shown to react equally well in the immunoassay (data not shown). IDA, however, allows determination of each form of the C-peptide specifically, since the 33-residue C-peptide elutes 2 min after the 31-residue C-peptide by microbore HPLC and they exhibit a distinct difference in mass.
The concentration of the 33-residue C-peptide determined in four of the samples by IDA was 80, 130, 230, and 470 pM, where the 31-residue C-peptide was determined at a concentration of 2100, 2540, 3320, and 4600 pM, respectively. Thus the 33-residue C-peptide concentration in these samples ranged from 4 to 10% of the 31-residue C-peptide concentration. The amount of the 33-residue C-peptide was too low to be measured in the remaining six samples, where the 31-residue C-peptide concentration was less than 1600 pM.
Measurement of Basal Levels
The insulin, C-peptide, and proinsulin concentrations of each
sample as determined by IDA were compared with immunoassay results (Figs. 5, 6, 7). Insulin and
C-peptide basal level values by IDA are at the lower end of the
standard values determined by immunoassay (7, 8, 33). Samples from
obese individuals show protein levels that were between those from
non-diabetic non-obese and type II diabetic individuals.
In most cases, insulin and C-peptide levels are higher by immunoassay than by IDA. Interestingly, the difference between MEIA and IDA results is highest for type II diabetic samples, where insulin concentrations are higher (Table II). Although results by IDA show that insulin, C-peptide, and proinsulin levels are all increased in obese and type II diabetic individuals (by 4, 3, and 6 times, respectively), the ratio between proinsulin and insulin levels is increased, whereas the ratio between C-peptide and insulin levels is decreased relative to non-obese non-diabetic individuals (Table II).
|
Post-pancreatic Graft Studies
Patient 1The insulin and C-peptide concentration of
post-operation samples taken over the 60-h period are slightly lower by
IDA than by immunoassay (Fig. 8). Both sets of results
remain close to standard basal level values although there is a slight
decrease in their concentration over the period. Proinsulin levels are extremely high, fluctuating between 60 and 170 pM
(approximately 10 times the normal basal concentration).
Patient 2
Following the administration of glucose to this
patient after an operation, insulin and C-peptide levels showed a
normal increase to maximum in 20 min followed by a fast decrease to
stability after 30 min (Fig. 9). Results by IDA are
similar to those by immunoassay although the C-peptide concentration is
lower by IDA during the initial 20 min. Proinsulin levels follow a
similar trace but increase faster, with a maximum after 10 min.
The difficulty in synthesizing analogues of macromolecules has often hindered studies using isotopic dilution methodology. For these investigations, analogues of insulin, C-peptide, and proinsulin were successfully prepared, possibly representing the largest target molecules to have been successfully subjected to the technique of stable isotope dilution mass spectrometry. Neither of the isotopes used in the preparation of these analogues, 15N or 2H, are radioactive or harmful in low doses. Thus, each of the isotope-labeled analogues used for initial in vitro work stands as an ideal candidate for possible in vivo work, where exogenous and endogenous protein would be distinguishable by ESI-MS.
The overall protein concentration of serum is too high for it to be feasible to introduce unfractionated samples directly into a mass spectrometer or an HPLC (where even in the presence of a carrier, target proteins are very difficult to detect against the large UV absorbance of serum). A specific extraction methodology is thus required for the quantitation of single proteins. Extraction by immunoaffinity is most specific when the antibody used reacts solely with the target protein. This is often not the case, however, leading to inaccuracies in techniques relying solely on immunological properties. Insulin antibodies used for this study by IDA were shown to bind insulin almost specifically with low C-peptide or proinsulin cross-reactivity, but the C-peptide antibodies bind both C-peptide and proinsulin with equally high affinity. Our method takes advantage of immunological cross-reactivity by using a mixture of antibodies against insulin and C-peptide to extract all three proteins, which are then easily separated and purified by chromatography. Alternatively, solid-phase methodology is far more adaptable to the extraction of many different proteins since antibodies are not required. Although this extraction technique is less specific, the carrier allows clear identification of the target protein from the low quantity of remaining impurities by subsequent chromatography.
Comparison of IDA against standard immunoassay serves to validate our methodology and indicate potential differences. Results were higher by immunoassay than by IDA for many of the samples studied. Differences between the two sets of results cannot be explained purely by experimental error or a difference in the concentration of standards used by each technique. However, the difference between determination of the C-peptide by CEI (which measures the sum of 31- and 33-residue C-peptide levels indistinguishably) and IDA (where we measure the 31-residue C-peptide levels specifically) may be partially explained by the significant proportion of the 33-residue C-peptide present in serum.
It is now widely accepted that insulin and C-peptide antibodies used in immunoassay techniques can cross-react with proinsulin and its conversion intermediates (4, 5). Although we measured low proinsulin reactivity (not more than 1%) in both MEIA and CEI, these immunoassays were shown to cross-react linearly with 56% (MEIA) and 42% (CEI) of des-31-32 split proinsulin when added in high purity to a series of test samples at known concentrations (from 20 pM to 1 nM) (data not shown). The preparation of des-31-32 split proinsulin will be described elsewhere.2 Since des-31-32 split proinsulin has been found to be the major conversion intermediate, present at up to 30% of the total insulin-related material in the circulation (4, 7, 34), insulin and C-peptide levels may be overestimated by MEIA and CEI in some cases.
In type II (and pre-type I) diabetic individuals, high levels of proinsulin and its conversion intermediates have been previously found relative to insulin (4, 13, 23-26, 35). However, this is only a relative measurement. Most conventional techniques cannot measure the absolute change in concentration of each protein specifically. Our studies indicate that, although insulin levels are increased in these individuals, abnormality in the ratio of circulating components is primarily caused by a large increase in proinsulin levels (Table II). This may be due to overstimulation of the pancreatic beta cells beyond their capacity to complete the proinsulin-to-insulin processing (possibly caused by insulin resistance or decreased glucose tolerance) or an intrinsic defect in expression or structure of the conversion enzymes (36, 37).
Studies using IRMA (a monoclonal antibody-based two-site immunoradiometric assay) suggest that both proinsulin and des-31-32 split proinsulin levels increase in type II diabetic subjects (4, 7, 8). Although IRMA is one of the most specific immunoassays to date, the proinsulin conversion intermediates cannot yet be fully distinguished. Since IDA does not require antibodies of high specificity, we feel that the individual measurement of each conversion intermediate would be a promising area of research using this new methodology.
A disproportional increase in the levels of des-31-32 split proinsulin would increase the proportion of cross-reactivity measured by MEIA. Our results, therefore, may confirm studies by IRMA since we see a larger difference between MEIA readings and the more specific IDA readings of insulin in samples from type II diabetic subjects (Table II). Notably, the ratio between MEIA and IDA values for insulin levels in non-diabetic individuals was calculated to be 1.3 (Table II). This is in good agreement with our preliminary comparison of the two techniques by measurement of insulin levels in samples taken from various types of individuals who were mainly non-diabetic, where the ratio between MEIA and IDA results was 1.2 (Fig. 3B).
Results by IDA indicate that C-peptide levels increase in obese and type II diabetic individuals to a lower extent than insulin levels do, resulting in a significant decrease in the ratio between them (Table II). The exact role of the C-peptide is not clearly understood although it exhibits specific biological activity of its own, with a partial effect on glucose transport, blood flow, capillary diffusion capacity, glucose utilization, and renal function in type I diabetic patients (38, 39). Insulin and the C-peptide are secreted in an equimolar ratio after the enzymatic breakdown of proinsulin. Since insulin is prone to a variable hepatic extraction and peripheral clearance (40, 41), C-peptide levels are frequently used to measure insulin secretion and beta cell function. However, this measure may not be accurate since our studies suggest that the ratio between insulin and C-peptide levels is variable between different types of individuals (possibly due to variability in the rate of elimination of both insulin and C-peptide from the circulation).
The post-pancreatic graft studies represent examples of a clinical situation where the accurate measurement of protein levels is of critical importance in assessing the success of the operation. Results obtained by IDA compare well with immunoassay data. The high and unstable level of proinsulin measured in patient 1 immediately after the operation was thought to be due to initial overcompensation of the secretory activity of the pancreatic beta cells following a decrease in their mass after autotransplantation. Several days later, proinsulin levels returned to normal (data not shown). This phenomenon was not detected before investigations by IDA, since insulin and C-peptide levels are relatively normal. The slow decrease in insulin and C-peptide concentration may simply be restabilization of the pancreas. Patient 2 showed a normal response to glucose, but the high initial increase in proinsulin levels detected by IDA could have resulted from overcompensation of the beta cells as discussed for patient 1.
We conclude that, in general, results between immunoassay and IDA correlate well. However, where conventional immunoassays (such as routinely used forms of radioimmunoassay and MEIA) are perturbed by such factors as very high or very low concentration levels (where calibration strays from linearity) or the presence of extraneous cross-reacting material (the magnitude of which cannot be calculated), much larger deviations are seen. Significant differences between the two sets of results outline the need of a definitive technique that avoids the problems associated with present methodology, thus increasing the accuracy of important clinical results and their interpretation.
Specifically, these studies demonstrate how IDA could be used toward understanding the behavior of insulin in humans. A precise, simultaneous measure of all insulin-related proteins in blood, including proinsulin conversion intermediates and insulin fragments, would help answer many of the questions relating to diabetes and become a valuable diagnostic and prognostic test in itself. Abnormal ratios between insulin and proinsulin-related proteins may, for example, be a good index of a pancreatic transplant on the verge of failing. Current methods, when the whole organ is involved, are based on determination of amylase in urine. Even this rather indirect method is not possible when the islets alone are transplanted.
Improvement in the accuracy of measurement of protein levels in humans is an important step toward a better understanding of many common and serious diseases. IDA is a conceptually simple and adaptable technique that could be used to study many different metabolic or physiological processes. Although for clinical analyses IDA is not as routine a method as standard immunoassay since measurements take longer to perform and special equipment is required, it could provide additional insight into difficult or suspect cases. Continual improvements in protein labeling facilities and mass spectrometric instrumentation could allow IDA to become a powerful tool for general studies on protein behavior.
We thank Drs. Alain Golay and Yves Morel for providing clinical serum samples, Dr. Jan Markussen of Novo Nordisk, Denmark, for providing antiserum against human C-peptide (mouse IgG-1), Prof. Philippe Halban and Yves Brandenburger for important discussions, Drs. Reto Stöcklin, Jean-François Arrighi, and Kahn Hoang-Van Fol Wolk on the modification of the proinsulin expression system, and Lis Campos for technical assistance.