Clinical Chemistry 47: 316-321, 2001;
(Clinical Chemistry. 2001;47:316-321.)
© 2001 American Association for Clinical Chemistry, Inc.
Long-Term Evaluation of Electrospray Ionization Mass Spectrometric Analysis of Glycated Hemoglobin
Norman B. Roberts1,a,
Alieu B. Amara1,
Michael Morris2 and
Brian N. Green2
1
Department of Clinical Chemistry, Royal Liverpool University Hospital, Liverpool L7 8XP, United Kingdom.
2
Micromass (UK) Ltd., Wythenshawe, Manchester M239LZ,
United Kingdom.
a Author for correspondence. Fax 44-0151-706-5813; e-mail
n.b.roberts{at}liverpool.ac.uk.
 |
Abstract
|
|---|
Background: Electrospray ionization mass spectrometry (ESIMS) has
been successfully applied to the identification of hemoglobin (Hb)
variants and the presence of glucose adducts (mass difference of 162
Da) on the separate Hb 
and ß chains. To establish the potential
of ESIMS as a routine and/or a reference method for the quantification
of glycohemoglobin (HbA1c), we carried out a detailed evaluation over a
4-month period in a routine laboratory environment.
Methods: We optimized a procedure using ESIMS suitable for the
routine quantitative analysis of HbA1c. We determined reliability and
reproducibility over 4 months and assessed the potential for automated
sample injection. We then compared values of 1022 blood samples from
diabetic patients with a routine HPLC-based ion-exchange procedure
(HA-8140; Menarini).
Results: Results of HbA1c measurement by ESIMS were available
within 3 min. The analytical imprecision (CV) was 1.6–5.0% for
both manual and automated injections. Data collection over the
m/z 980-1400 range confirmed lower glycation of the
chain relative to the ß chain (0.66:1). Only one glycation was
observed per globin chain. The overall glycohemoglobin (i.e., the
average of - and ß-chain glycations) measured by ESIMS
(x) on 1022 blood samples was lower than by HPLC
(y): y = 1.0432x +
0.4815. However, the ß-chain glycation measured by ESIMS was
up to 20% higher than the value measured by ion-exchange HPLC and
showed a close conformity, particularly at 5–10% HbA1c, with the
ion-exchange Diabetes Control and Complications Trial (DCCT)-corrected
and the United Kingdom National External Quality Assessment
Scheme DCCT mean return values.
Conclusions: ESIMS provides a precise measurement of HbA1c and,
in particular, glycation of the ß chain. The method is robust and
could be proposed as a procedure to substantiate HbA1c measurement
and/or calibration.
|
Introduction
|
|---|
There are several different ways of measuring glycohemoglobin
(HbA1c);1
the longer term (months) glucose control marker (1)
uses ion-exchange and affinity chromatography, electrophoresis, and/or
immunologically based assays. Unfortunately, these separate procedures
produce different analytical values (2)(3). It
is therefore widely recognized that both a reference method and a
reference standard are required. To date, most assays report values in
Diabetes Control and Complications Trial (DCCT) "equivalents" based
on an ion-exchange procedure, the designated comparison method
(2). Although actual measurement of the glycation of the
N-terminal valine of the ß chain of hemoglobin (the fraction defined
as HbA1c) is possible, this is not yet available as a routine procedure
(4).
Electrospray ionization mass spectrometry (ESIMS) has been successfully
applied to the identification of hemoglobin (Hb) variants
(5) and the measurement of glucose adducts on the and
ß chains. A mass increase of 162 Da is observed incremental to the
native protein chain and is assigned as the addition of
glucose (mass, 180 Da) through elimination of water (-18 Da). ESIMS
has been successfully applied to the quantification of this glucose
addition to Hb as a measure of HbA1c (6). However,
establishing the potential of ESIMS as a routine and/or a reference
method for HbA1c requires a more detailed evaluation over a longer
period in a routine laboratory environment (7).
We therefore assessed the reliability and precision of the measurement
of HbA1c by ESIMS, in particular its potential for total automation, in
a study that lasted several months.
|
Materials and Methods
|
|---|
chemicals
All of the chemicals used were Analar grade (Sigma).
Acetonitrile was HPLC grade (BDH Laboratory Supplies), and water was
deionized grade (Elga Ltd). Quality-control samples used were
Lyphocheck Whole Blood Diabetic Controls 1 and 2 with stated HbA1c
values (ion-exchange chromatography) of 4.8% ± 1.0% and 8.8% ±
1.2% (mean ± SD), obtained from Bio-Rad Laboratories Ltd.
Blood samples were obtained from patients (n = 1022) with diabetes
mellitus attending the Royal Liverpool University and Broadgreen
Hospitals over a 4-month period. The routine analysis of these samples
for HbA1c was part of the assessment of the patients’ glycemic control
and was carried out using a HI-AUTO, HA-8140 automated analyzer
(Menarini). To correct ion-exchange values (x) to DCCT
equivalents (y), a correction factor, y =
1.02x +0.87 (supplied by Menarini using a certified method
comparison), was used. Blood analysis by ESIMS was carried out
anonymously and in accordance with the ethical regulations of the Royal
Liverpool University Hospital.
mass spectrometric analysis
Whole blood, always less than 3 days old, was diluted 10-fold in
deionized water and a further 50-fold in aqueous acetonitrile to yield
a final aqueous solution containing 500 mL/L acetonitrile and 2
mL/L formic acid to form the working solutions. The working
solutions were then introduced into a single quadrupole Platform LC
mass spectrometer (Micromass) at flow rates of 
20 µL/min in 500
mL/L aqueous acetonitrile mobile phase. The instrument was run in the
positive ion electrospray (ES+) mode, and a 3.5 kV potential was
applied to the spraying capillary tip. The spray was desolvated using
nitrogen gas at a source temperature of 80 °C. Eleven 5-s scans were
summed over a m/z range of 980-1400 with an interscan
time of 0.1 s and a total analysis time of 1 min. The mass scale
of the instrument was externally calibrated using horse heart myoglobin
(mass, 16 951.1 Da), and the acquired mass spectra were individually
internally calibrated using human chain (15126.4 Da) typically
present in the samples, as described previously (6).
Automated injection of the diluted samples was also carried out using a
Waters Alliance 2690 autosampler module (Waters UK Ltd). The flow
program of the solvent management system allowed for an interinjection
time of 4 min, with an observed carryover from sample to sample of
<1%.
The raw data were processed using the MassLynx data processing
algorithms provided with the instrument. The signal-to-noise ratio
observed for the raw data was typically in excess of 50 for the base
peak in the spectrum [at m/z 1009.4 for
(+15H)15+ or m/z 992.7 for
(ß+16H)16+]; the noise was measured at a point
where no sample-related signals were observed (e.g., m/z
1280–1310). The algorithm used to transform the data was the MassLynx
Transform algorithm. Although the routines are not explicitly named
(Background, Subtract, Centroid, and Transform), their use is implicit
in the text. Thus, after baseline subtraction, the raw data were
transformed to a true molecular-weight scale and smoothed, and the
heights of the peaks were used as a measure of their intensities. The
ion intensities observed under electrospray ionization conditions
are concentration dependent. Relative concentrations may therefore be
measured by comparing the ion signals observed for the compounds of
interest. However, the relative sensitivities of the and non-
species might be expected to be slightly different because of the
different compositions of the proteins. This was countered by tuning
the conditions as outlined such that the signal intensities of the
-chains [(+15H)15+, m/z
1009.4] and ß-chains [(ß+16H)16+,
m/z 992.7] were within 10% of each other. Any analyses
outside these limits were discarded and the samples reanalyzed. This
was checked by analyzing quality-control samples at the beginning,
middle, and end of the daily run and to provide three daily sets of
control values.
calculation of percentage of glycation
The principles of the calculation of the percentage of glycation
have been described previously (6). In the absence of direct
measurements of the sensitivities of many of the species apparent in
the electrospray mass spectrum, the following assumptions were made to
derive an expression to allow the percentage of glycation to be
calculated from the experimental data:
- All -chain species, including noncovalent adducts, have
the same sensitivity (same mass spectrometer response for a given molar
concentration in the analyte). We believe that this is a reasonable
assumption to make because these species are the same protein chain
with relatively minor modifications.
- All ß-chain species, including noncovalent adducts, have the
same sensitivity.
- The sensitivities of the - and non--chain species are
different (set to within 10%) and are related by: sensitivity of
-chain species = K x sensitivity of
non--chain species (where K is constant for a given set
of operating conditions).
- The relative intensities of adducted glycated -chain to
glycated -chain species are the same as those of adducted -chain
to -chain species. Similar assumptions are made for the
corresponding ß-chain species.
- Minor non--chain species may be ignored. The most prevalent
other non- species present in apparently healthy adults is
the
chain, which is 2% of the ß chain. However, at this
concentration, glycation of the chain would be undetectable. On
this basis, we chose to ignore any contribution from minor non-
species.
- There is no contribution to the glycated species by multiple
alkali metal adducts of the nonglycated species; desalting the sample
does not affect the overall measurement (6).
On the basis of these assumptions and the knowledge that the ESIMS
analyte originates from whole-blood Hb, in which there are equal molar
amounts of - and non -chain species, it can be rigorously shown
that:
where GHb is glycohemoglobin, and ß represent the
intensities of the and ß chains, and g
and ßg represent the intensities of the
glycated and ß chains. This simple expression means that only
four peaks need to be measured from the transformed spectrum, but it is
based on the assumptions outlined, many of which are difficult to prove
empirically. Nevertheless, the close correlation between the ESIMS
results and the results obtained with other techniques imply that the
assumptions are not unreasonable.
Data comparisons were carried out using standard regression confirmed
by Passing and Bablok analysis and Bland-Altman comparative analysis.
|
Results
|
|---|
A typical transformed mass spectrum is shown in Fig. 1
, indicating the and ß chains and the glycated fractions.
Only one glucose addition was seen at the respective chain mass + 162
Da, i.e., normal chain at 15 288 Da and normal ß chain at
16 029 Da, with no peak at mass + 324 Da (equivalent to two glucose
molecules) at any HbA1c value in this study. A comparison of the -
and ß-chain glycation with increasing HbA1c values confirmed the
increased glycation of the ß chain relative to the chain
(ßg = 1.43g + 0.87;
Fig. 2A
). The difference between the two values increased consistently
as glycation increased (Fig. 2B
), with -glycation always 66% of
the ß-glycation.
The imprecision analysis in Table 1
shows good analytical performance during a typical working day
(intraassay variation) and over the 4-month study period
(interassay variation), with values of 1.6–5.0% for
both manual and automated sample processing.
Comparison of the overall glycation determined by ESIMS and an
ion-exchange chromatographic procedure showed good agreement (Fig. 3
). The graphical statistical analysis indicated that values for
HbA1c were on average 0.7% lower with ESIMS. Comparison of ESIMS
ß-glycation with DCCT-corrected ion-exchange values gave good
agreement, particularly for 5–10% HbA1c (Fig. 4
) with ESIMS showing an overall lower value of mean 0.4%.
View larger version (23K):
[in this window]
[in a new window]
|
Figure 3. Comparison of percentage of total glycation as measured by
ESIMS vs ion-exchange chromatography (A) and
Bland-Altman plot showing the differences between the methods
(B).
|
|
View larger version (25K):
[in this window]
[in a new window]
|
Figure 4. Comparison of percentage of Hbß glycation, as measured
by ESIMS, with total glycation, as calculated using DCCT-corrected ion
exchange results (A) and Bland-Altman plot showing the
differences between the methods (B).
|
|
The ß-glycation values determined by ESIMS (y) show good
agreement (Fig. 5
) with the mean [United Kingdom National External Quality
Assessment Scheme (UKNEQAS)] DCCT-aligned values (x) of
5.8%–12.0%. The correlation r2 was
0.96 (P <0.001) and the equation for the line was:
y = 0.7825x + 1.4761 (n = 23).
The ESIMS values were consistently higher at glycation values >10%.
View larger version (12K):
[in this window]
[in a new window]
|
Figure 5. Comparison of percentage of Hbß glycation, as measured
by ESIMS, and the mean DCCT-aligned HbA1c values from UKNEQAS
(A) and Bland-Altman plot showing the differences
between the two methods (B).
|
|
|
Discussion
|
|---|
The data presented on the measurement of HbA1c confirm the overall
reliability and reproducibility of ESIMS, thus satisfying the various
criteria suggested (7) to substantiate the possibility of
using this technique as a routine procedure. Indeed, the technique is
well established as routine in the pharmaceutical industry and is now
recommended for the complex analysis of amino acid abnormalities in the
newborn (8). However, for reliable glycated Hb analysis, the
standard operating procedures we have outlined (6) must be
strictly adhered to. Studies (9) where mass spectra were
acquired over the m/z 600-1400 range showed substantial
reduction in the signal intensity for ß chain (particularly with the
higher m/z ratios). In these reports, ESIMS was used to
identify components obtained during chromatography, i.e., in a
qualitative mode. Accordingly, if the intensities of the and ß
chains are not within 10% of each other, then the measured ß
glycation will be grossly overestimated and an inappropriate value
obtained. We have also confirmed that -chain glycation is
consistently less than ß-chain glycation. The explanation for this
needs further study to establish whether these differences are related
to site-specific glycation, i.e., whether the N-terminal
NH2 of the chain is the only site of
glycation and, therefore, whether the other possible sites on the
chain are less accessible than those on the ß chain.
Interestingly, the percentage of glycation of the ß chain showed much
closer agreement with the DCCT-corrected ion-exchange values, in
particular, at 5–10% HbA1c. This would infer that the DCCT
procedures are measuring the glycated fraction of the ß chain. A
similarity between an ESIMS-based assay for ß chain glycation and
ion-exchange chromatography for HbA1c values, in particular for
concentrations between 6% and 10%, has also been observed
(4). These authors argued that the poor agreement obtained
at <5% HbA1c was probably attributable to impurities in the
chromatographic peak causing higher values in high-performance
ion-exchange chromatography, whereas the lower values obtained
at >12% HbA1c suggested additional glycation of the 
-amine groups
of lysine. This latter explanation, however, seems unlikely because
this would be associated with a doubly glycated ß chain, which we
have never observed even at a high percentage of glycation (18%). It
may be that higher glycation values somehow affect column separation
efficiency, causing more HbA1c to co-elute with HbAo. The possibility
that more than one glycation occurs at a detectable concentration was
again not substantiated by our findings. Indeed, applying the following
argument suggests that double glycation will be always low. If we
assume that there two sites for glycation, then the possibility of
seeing both sites occupied on one molecule can be expressed by the
following binomial distribution. Suppose the two glycation sites, e.g.,
on the ß chain, can randomly attach a glucose moiety with
probabilities p1 on site 1 and p2 on site 2. The
proportions of no glycation (0G) to one glycation (1G) to two
glycations (2G) are (1 - p1)(1 - p2)
to p1(1 - p2) + p2(1 -
p1) to p1p2. The sum of these proportions is 1.
The highest proportion of two glycations should occur when two sites
have equal probability of being occupied, i.e., when p1
= p2, simplified as p. In this case, the
probabilities of obtaining 0G:1G:2G are (1 -
p)2:2p(1 -
p):p2. The highest
proportion of one glycation, 1G/(0G + 1G), observed to date by ESIMS
was 0.18, and in the above equation, the value of p that
gives 0.18 on the ß chain is 0.0989. Thus,
p2, the predicted proportion of double
glycation, is 0.0098, or 1.0%. If the second site has a lower
probability of being occupied than the first, then the fraction of
double glycations on the ß chain will be reduced. Thus, even in the
most favored conditions for two glycations occurring simultaneously,
only 1% of the molecules will be doubly glycated even at the high
single glycation of 18% on the ß chain. To date, we have not been
able to reliably detect double glycations at this percentage.
The analysis we have outlined measures the overall percentage of
glycation of the ß chain and not the percentage of glycation of the
N-terminal valine. This is unlike the method proposed as a reference
procedure for HbA1c, where the actual terminal glycated cleavage
product from the ß chain (as confirmed by ESIMS) is measured
(4). It is nevertheless interesting that there is such close
agreement between our procedure for whole-blood analysis of the
percentage of ß-chain glycation by ESIMS and the DCCT-aligned values
obtained by high-performance ion-exchange chromatography. This would
imply that validation of DCCT alignment could be confirmed by direct
analysis of whole blood using ESIMS. If further analytical validation
was required, comparative analysis with washed cells or even partially
purified hemoglobin from cells could be carried out and a certified
reference material produced. The question then arises as to how
accurate is measurement by mass spectrometry. Clearly, instrumental
conditions for the mass spectrometer must be properly defined to ensure
that the signal intensities of the and ß chains are similar or
the analysis will be only qualitative (9). Quantitative
agreement between whole-blood analysis of ß-chain glycation by ESIMS
and a DCCT-aligned chromatographic procedure for HbA1c was confirmed
using data collected over the m/z 980-1300 range
(10). In addition, independent observations (11)
using similar conditions have confirmed the reliability and robustness
of ESIMS for measurement of HbA1c, whereas poorer agreement was
observed using data collection over a larger range, m/z
700-1400 (12).
The notion that the ESIMS technique is too sophisticated and/or
expensive for most routine clinical laboratories may be an erroneous
supposition. Indeed, although the overall capital cost is high, the
actual running cost of reagents is very low (a few pence/cents), which
means that the overall cost per test would be little different from
current methodologies. Certainly the speed of analysis is not an issue
because with automated analysis and calculation of data, the result can
be available within 3 min. Measures are currently being investigated to
reduce the cycle time, and analysis times approaching 2 min have been
demonstrated after manual injection. However, the conditions for
automation would need to be optimized further because initial data on
analytical precision were relatively worse (although just acceptable)
after automated compared with manual injection. Other benefits of ESIMS
include the identification of Hb variants, although the classification
of the amino acid alteration requires further detailed analysis
(13).
In conclusion, we have confirmed that accurate and reliable
quantification of HbA1c is possible by ESIMS in a routine environment.
The glycation of the and ß chains enables an overall HbA1c index
to be measured. The separate ß-chain glycation values obtained by
ESIMS show good agreement with DCCT-aligned ion-exchange chromatography
values, inferring that these different techniques are measuring a
common analyte. Thus, the mass spectrometric technique, operated under
the appropriate conditions, may be suitable as a reference method for
ß glycation and thus for DCCT alignment, particularly if previously
used methods become unavailable.
|
Footnotes
|
|---|
1 Nonstandard abbreviations: HbA1c, glycohemoglobin; DCCT, Diabetes Control and Complications Trial; ESIMS, electrospray ionization mass spectrometry; Hb, hemoglobin; and UKNEQAS, United Kingdom National External Quality Assessment Scheme. 
|
References
|
|---|
-
. DCCT Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin dependent diabetes mellitus. N Engl J Med 1993;329:977-998.[Abstract/Free Full Text]
-
John WG. Glycated haemoglobin analysis. Ann Clin Biochem 1997;34:17-31.
-
Weykamp CW, Penders TJ, Muskiet FAJ, Van der Slik W. Effect of calibration on dispersion of glycohemoglobin values determined by 111 laboratories using 21 methods. Clin Chem 1994;40:138-144.[Abstract/Free Full Text]
-
Kobold U, Jeppsson J-O, Dulffer T, Fuike A, Hoelzel W, Miedema K. Candidate reference methods for hemoglobin A1c based on peptide mapping. Clin Chem 1997;43:1944-1951.[Abstract/Free Full Text]
-
Shackleton CHL, Witkowska HE. Mass spectrometry in the
characterization of variant hemoglobins. In: Desiderio DM, ed. Mass
spectrometry: clinical and biomedical applications, Vol. 2. New York:
Plenum, 1994:135–99..
-
Roberts NB, Green BN, Morris M. Potential of electrospray mass spectrometry for quantifying glycohemoglobin. Clin Chem 1997;43:771-778.[Abstract/Free Full Text]
-
Miedema K. Electrospray mass spectrometry for measurement of glycohemoglobin. Clin Chem 1997;43:705-707.[Free Full Text]
-
Rashed MS, Bucknall MP, Little D, Awad A, Jacob M, Alamandi M, et al. Screening blood samples for inborn errors of metabolism by electrospray tandem mass spectrometry with a microplate batch process and a computer algorithm for automated flagging of abnormal profiles. Clin Chem 1997;43:1129-1141.[Abstract/Free Full Text]
-
Peterson KP, Pavlovich JG, Goldstein D, Little R, England J, Peterson CM. What is hemoglobin A1c? An analysis of glycated hemoglobin by electrospray ionization mass spectrometry. Clin Chem 1998;44:1951-1958.[Abstract/Free Full Text]
-
Rocks BF, Morris M, Cooper D, Iversen SA. Automated
measurement of glycated hemoglobin by mass spectrometry. In: Proc ACB
National Meeting. London: Association of Clinical Biochemists,
1999:112..
-
Nakanishi T, Mizaki A, Kishikawa M, Yasuda M, Tokuchi Y, Kanada Y, Shimizu A. Quantification of glycated hemoglobin by electrospray ionization mass spectrometry. J Mass Spectrom 1997;32:773-778.[Web of Science][Medline]
[Order article via Infotrieve]
-
Lapolla A, Fedele D, Plebani M, Garbeglio M, Seraglia R, D’Alpioso M, et al. Direct evaluation of glycated evaluation of glycated and glyco-oxidized globins by matrix-assisted laser desorption/ionization mass spectrometry. Rapid Commun Mass Spectrom 1999;13:8-14.[Web of Science][Medline]
[Order article via Infotrieve]
-
Witkowska HE, Bitsch F, Shackleton CH. Expediting rare variant hemoglobin characterization by combined HPLC/electrospray mass spectrometry. Hemoglobin 1993;17:227-242.[Web of Science][Medline]
[Order article via Infotrieve]
The following articles in journals at HighWire Press have cited this article:
|
|
|
|
 
N. Ahmed, R. Babaei-Jadidi, S. K. Howell, P. J. Thornalley, and P. J. Beisswenger
Glycated and Oxidized Protein Degradation Products Are Indicators of Fasting and Postprandial Hyperglycemia in Diabetes
Diabetes Care,
October 1, 2005;
28(10):
2465 - 2471.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
