Perspectives in Diabetes
Role of Oxidative Stress in Development
of Complications in Diabetes
JOHN W. BAYNES
Nc-(carboxymethyl)lysine, Nc-(carboxymethyl)hydroxy-
lysine, and the fluorescent cross-link pentosidine
are formed by sequential glycation and oxidation
reactions between reducing sugars and proteins.
These compounds, termed glycoxidation products,
accumulate in tissue collagen with age and at an
accelerated rate in diabetes. Although glycoxidation
products are present in only trace concentrations,
even in diabetic collagen, studies on glycation and
oxidation of model proteins in vitro suggest that these
products are biomarkers of more extensive underlying
glycative and oxidative damage to the protein.
Possible sources of oxidative stress and damage to
proteins in diabetes include free radicals generated by
autoxidation reactions of sugars and sugar adducts to
protein and by autoxidation of unsaturated lipids in
plasma and membrane proteins. The oxidative stress
may be amplified by a continuing cycle of metabolic
stress, tissue damage, and cell death, leading to
increased free radical production and compromised
free radical inhibitory and scavenger systems, which
further exacerbate the oxidative stress. Structural
characterization of the cross-links and other products
accumulating in collagen in diabetes is needed to gain
a better understanding of the relationship between
oxidative stress and the development of complications
in diabetes. Such studies may lead to therapeutic
approaches for limiting the damage from glycation
and oxidation reactions and for complementing
existing therapy for treatment of the complications
of diabetes. Diabetes 40:405-12,1991
O
xidative stress may be defined as a measure of
the steady-state level of reactive oxygen or ox-
ygen radicals in a biological system. A hypo-
thetical sequence of events by which oxidative
stress may be linked to tissue damage and the development
of pathophysiology is outlined in Fig. 1. According to this
scheme, increased oxidative stress may .result from over-
production of precursors to reactive oxygen radicals and/
DIABETES, VOL. 40, APRIL 1991
or decreased efficiency of inhibitory and scavenger systems.
The stress then may be amplified and propagated by an
autocatalytic cycle of metabolic stress, tissue damage, and
cell death, leading to a simultaneous increase in free radical
production and compromised inhibitory and scavenger
mechanisms, which further exacerbate the oxidative stress.
For practical reasons, neither the rate of oxidant produc-
tion nor the steady-state levels of reactive oxygen species
are easily measured in biological systems. Thus, oxidative
stress must be inferred from measurements of oxidative
damage as estimated from the kinetics of formation, the
steady-state levels, or the extent of accumulation of oxidation
products in tissues, plasma, or urine. However, the detection
of increased levels of oxidation products in tissues is not,
per se, sufficient to implicate oxidative stress in the pathology
unless the damage can be logically and quantitatively re-
lated to the development of pathology and until it can be
shown that inhibition of oxidative damage prevents or retards
the disease process.
The concept I develop in this article is that oxidative stress
may be a common pathway linking diverse mechanisms for
the pathogenesis of complications in diabetes. Mechanisms
that contribute to increased oxidative stress in diabetes may
include not only increased nonenzymatic glycosylation (gly-
cation) and autoxidative glycosylation but also metabolic
stress resulting from changes in energy metabolism, alter-
ations in sorbitol pathway activity, changes in the level of
inflammatory mediators and the status of antioxidant defense
systems, and localized tissue damage resulting from hypoxia
and ischemic reperfusion injury. The goal of this article is to
focus more on the common pathway, the role of oxidative
stress and damage in the development of complications,
rather than on the array of contributory mechanisms. For
From the Department of Chemistry and School of Medicine, University of South
Carolina, Columbia, South Carolina.
Address correspondence to J.W. Baynes, Department of Chemistry, Uni-
versity of South Carolina, Columbia, SC 29208.
Received for publication 6 December 1990 and accepted in revised form
11 December 1990.
405
OXIDATIVE STRESS IN DIABETES

ENZYMATIC REACTIONS Lipids Oxidation

Electron Transport
Oxidases Proteins Fragmentation
Men+
Oxygenases
Nucleic Acids Crosslinking
AUTOXIDATION
Metal catalyzed Glycoconjugates Fluorescence

111
FORMATION OF ACTION OF INHIBITORS
TARGETS DAMAGE
PRECURSORS AND SCAVENGERS

FIG. 1. General pathway by which increased oxidative stress may contribute to development of complications in diabetes. Representative
enzymatic and nonenzymatic sources of reactive oxygen are shown. Intermediates such as superoxide, hydrogen peroxide, and lipid peroxides
are precursors to more reactive species such as hydroxyl radical. Inhibitors include enzymes such as superoxide dismutase, catalase, and
peroxidases, which limit accumulation of precursors. Proteins such as transferrin, ceruloplasmin, and albumin also function as inhibitors by
limiting concentration of free transition metal ions (Me n+ ), which are catalysts of oxidation reactions. Radical scavengers limit hydroxyl radical
damage by trapping reactive radicals in both hydrophiiic and lipophilic (membrane) environments. Water-soluble scavengers include ascorbic
acid (vitamin C), glutathione, and uric acid; lipid-soluble scavengers include tocopherol (vitamin E) and ubiquinol. Net flux of free radicals,
representing level of oxidative stress, causes chemical modification of biological molecules. Resulting damage may affect cell and tissue
functions, leading to pathology.

more background on the relevance of oxidation in biological
systems, see ret. 1. There are also excellent reviews on the
role of free radicals in the etiology of diabetes (2), on the
possible role of altered antioxidant defenses in the devel-
opment of complications (2-4), and on the role of oxidation
of plasma lipids and lipoproteins in the development of ath-
erosclerosis in diabetes (5). For further development of my
viewpoints on oxidative stress in diabetes, see ref. 6.
APPROACHING THE QUESTION
If we accept that the complications of diabetes are in some
way an indirect manifestation of metabolic stresses resulting
from altered insulin homeostasis and energy metabolism,
then the critical questions from the viewpoint of this per-
spective are: 7) Do these metabolic stresses lead to in-
creased oxidative stress in diabetes? 2) If so, is the resulting
structural and/or functional damage sufficient to induce the
development of complications? I approach these questions
in a reverse order by asking first, What is the nature of the
tissue changes and damage associated with the develop-
ment of complications in diabetes? and then, Is oxidative
stress a likely source of damage? Because of the limited
information on oxidative damage to nucleic acids and gly-
coconjugates (glycolipids, glycoproteins, glycosaminogly-
cans) in diabetes, the discussion will focus on oxidative mod-
ifications of proteins and lipids, with emphasis on the role of
modifications of collagen in the development of vascular and
basement membrane pathology in diabetes.
appears to be the critical factor limiting the accumulation of
oxygen radical damage. However, for long-lived unrepair-
able protein molecules such as the collagens, products of
oxygen radical reactions may accumulate with time, and
through alterations in protein structure and function, these
oxidation products may contribute to the development of
pathology. Long-lived proteins therefore constitute a unique
sensor for exposure to oxidative stress and provide a con-
venient source for identification of products formed during
oxidative modifications of proteins.
Modifications of long-lived extracellular proteins (e.g.,
crystallins, collagens, elastins, laminin, myelin sheath pro-
teins) and structural changes in tissues rich in these proteins
(lens, vascular wall, basement membranes) are associated
with the development of complications in diabetes such as
cataracts, microangiopathy, atherosclerosis, and nephrop-
athy. The chemical and physical changes characteristic of
collagen in diabetes are summarized in Table 1. The physical
changes in collagen are directly related to the underlying
chemical modifications of the protein. Similar changes, both
chemical and physical, develop gradually during the normal
aging of collagen, but the process appears to be accelerated
TABLE 1
Chemical and physical changes in collagen in diabetes
Chemical Physical

NATURE OF COLLAGEN MODIFICATIONS IN DIABETES Increased glycation Increased browning

Increased pentosidine
There is no evidence that once oxidative damage occurs it Increased carboxymethylation
may be reversed, for example, by chemical or enzymatic Increased cross-linking
reduction of the oxidized species back to the native form. Maturation of reducible
In the case of DNA, repair enzymes act by excision and cross-links
Resistance to enzymatic
replacement of the modified base or nucleotide. For proteins,
digestion
lipids, and RNA, the kinetics of turnover of the molecule
Increased fluorescence
Increased mechanical strength
Increased thermal stability
Decreased solubility
Decreased elasticity
Resistance to denaturants

406 DIABETES, VOL. 40, APRIL 1991


in diabetes, depending on the severity and duration of dis-
ease. Alterations in collagen synthesis and turnover also
occur, and the structural changes are accompanied by mor-
phological and functional alterations in collagen-rich tissues
in diabetes, such as the thickening of basement membranes,
altered vascular permeability, decreased joint mobility, and
impaired wound healing.
LACK OF DIRECT EVIDENCE FOR INCREASED OXIDATIVE
MODIFICATION OF COLLAGEN IN DIABETES
For two reasons, it is not possible to make a firm statement
about the significance of oxidation in the chemical modifi-
cation of collagen in diabetes: first, there is limited infor-
mation on the nature of the chemical changes that occur in
proteins exposed to oxidative stress; second, there is even
less information on the nature of the chemical changes that
occur in diabetic collagen. Carbonyl compounds are known
to be formed from amino acids during metal-catalyzed oxi-
dation of proteins in vitro and in vivo (7). However, carbonyl
compounds are unstable in biological systems and do not
accumulate; they may react with amines or be further oxi-
dized to carboxylic acids. Some stable oxidation products,
such as aspartate, produced on oxidation of histidine, are
indistinguishable from the natural amino acids in protein, so
that evidence of oxidative damage is not readily detected.
Other products, particularly those derived from tryptophan,
may be destroyed during hydrolytic work-up of the protein.
In those cases where unique and stable oxidation products
are formed, e.g., o-tyrosine or m-tyrosine by hydroxylation
of phenylalanine or dityrosine by oxidative dimerization of
tyrosine, there is no information on their rate of accumulation
or concentration in diabetic compared with control collagen.
In summary, there are few clearly defined stable chemical
markers of oxidative damage to proteins, and those that have
been characterized have not been measured or shown to
increase in collagen in diabetes. This problem is not unique
to diabetes because little is known about the oxidation of
proteins, even in pathologies in which oxidative stress is
considered to have a more definitive role, such as ather-
osclerosis and rheumatoid arthritis.
INDIRECT EVIDENCE FOR INCREASED OXIDATIVE
DAMAGE TO COLLAGEN IN DIABETES
Despite the lack of information on amino acid oxidation prod-
ucts, studies on glycation of proteins and Maillard reactions
of glycated protein have yielded indirect evidence for in-
creased oxidative modification of collagen in diabetes. Thus,
in addition to the Amadori adducts, fructoselysine (FL) and
fructosehydroxylysine, formed on glycation of lysine and
hydroxylysine residues in collagen, there are three carbo-
hydrate-derived oxidation products that are increased in
diabetic compared with age-matched nondiabetic colla-
gen: A/^carboxymethyOlysine (CML), A/e-(carboxymethyl)-
hydroxylysine (CMhL), and pentosidine (Fig. 2). CML and
CMhL are formed by oxidative cleavage of Amadori adducts
(8,9), whereas pentosidine is a fluorescent (excitation/emis-
sion 328/378 nm) cross-link formed between lysine and ar-
ginine residues in protein (10,11). All three of these com-
pounds are autoxidation products, i.e., formed in reactions
in which the oxidant is a reactive form of oxygen. Thus, the
formation of CML and CMhL from Amadori compounds is
DIABETES, VOL. 40, APRIL 1991
J.W. BAYNES
'NH-CH-CO** wNH-CH-CO**
(CH 2 ) 4 (CH 2 ) 2
NH CHOH HN r< >1
AH2 {H2 HN^N^^
COOH NH
CH 2
1
^*Arglnlne.M*
COOH
N-(carboxy- N -(carboxymethyl)-
methyl)lysine hydroxylysine Pentosidine
FIG. 2. Structures of Maillard reaction products known to accumulate
in collagen with age and at accelerated rate in diabetes. Because of
role of carbohydrate and oxygen in their formation, these compounds
have been termed glycoxidation products. They are considered
biomarkers of extent of glycative and oxidative damage to proteins.
inhibited under anaerobic conditions and by metal ion che-
lators and oxygen radical scavengers in aerobic systems
(8,12). The formation of pentosidine during browning of pro-
teins by sugars or during synthesis from carbohydrate and
amino acid precursors is also inhibited under anaerobic con-
ditions. Although the terms Maillard and browning reaction
are often used synonymously to describe events occurring
after the glycation reaction, CML and CMhL are colorless
and have been described as products of nonbrowning path-
ways of the Maillard reaction (12). In contrast, pentosidine
is a true browning product, with maximum absorbance at
328 nm, and its concentration correlates strongly with total
fluorescence in collagen, measured at either 328/378 or
370/440 nm. The presence of these oxidation products in
glycated collagen is not surprising, because it has been
known for decades that the Maillard reaction in vitro is stim-
ulated by oxygen and catalysts of oxidation reactions such
as phosphate and traces of transition metal ions and inhib-
ited by reducing agents such as ascorbate, bisulfite, and
thiol compounds (13).
The increased concentrations of CML, CMhL, and pen-
tosidine in diabetic collagen provide indirect evidence for a
diabetes-related increase in oxidative damage to the protein
(11). The evidence is indirect because it is not the amino
acids in protein that have become oxidized but rather the
attached carbohydrate. However, the assumption that the
increase in carbohydrate oxidation products signifies an in-
crease in underlying oxidative damage to the protein is reas-
onable because glycation of proteins in vitro may be ac-
companied by oxidative fragmentation of the protein and
peroxidation of associated lipids (4,14,15). Amadori adducts
are also a ready source of superoxide, e.g., in the fruc-
tosamine assay (16,17), providing experimental support for
the argument that glycation of protein enhances its potential
exposure to oxidative damage. Glycation also enhances the
development of fluorescence during oxidation of proteins
(18), and the wavelength maxima of fluorescence generated
during browning of proteins in vitro are similar if not indis-
tinguishable from that found in oxidized proteins (19,20).
Thus, it is possible that much of the increase in collagen-
linked fluorescence observed in diabetes is the result of a
glycation-dependent enhancement of autoxidative reac-
tions. This argument is supported graphically by the three-
dimensional (3-D) fluorescence spectra shown in Fig. 3. The
407
 OXIDATIVE STRESS IN DIABETES
360
360
similarities in the 3-D spectra of natural skin collagen and
browned RNase support the involvement of Maillard reac-
tions in the browning of protein in vivo (21; Fig. 3, A and 6).
The similarities between these spectra and those of RNase
oxidized by exposure to oxygen radicals generated either
by ionizing radiation (Fig. 3C) or metal-catalyzed oxidation
(Fig. 3D) suggest that much of the fluorescence in browned
proteins may be formed by secondary oxidation reactions.
Admittedly, the spectra are somewhat featureless, and the
structures of the fluorescent products are largely unknown,
but those products that are known to accumulate in collagen
in diabetes (CML, CMhL, and pentosidine) are, in fact, prod-
ucts of oxidation reactions, suggesting that other oxidation
products, both fluorescent and nonfluorescent, may also be
formed.
AUTOXIDATIVE GLYCOSYLATION AND GLYCOXIDATION
Wolff (4) introduced the term autoxidative glycosylation to
describe the proposed role of reducing sugars as catalysts
of the oxidative chemical modification and cross-linking of
proteins (14). Autoxidative glycosylation is initiated by the
oxidation of an aldose or ketose to a more reactive dicarbonyl
sugar (glucosone), which would then react with protein to
form a ketoimine adduct. This adduct is related to but more
reactive than the ketoamine adduct formed by the Amadori
rearrangement and would also initiate further Maillard or
browning reactions. The reduced oxygen products formed
in the autoxidation reaction include superoxide and hydro-
gen peroxide (16,17,22), which, in the presence of metal
ions, would cause oxidative damage to neighboring mole-
cules. Therefore, autoxidative glycosylation is a reasonable
mechanism for the production of free radicals, leading to
fragmentation of proteins (4,14) and oxidation of associated
lipids (14) during glycation reactions.
408
EMISSION
FIG. 3. Three-dimensional (3-D) fluorescence
spectra. A: pepsin-solubilized insoluble skin
collagen isolated from 50-yr-old diabetic patient;
B: bovine pancreatic RNase after incubation with
EMISSION 250 mM glucose in 0.2 M phosphate buffer, pH
7.4, for 28 days at 37°C under air; C: RNase
after oxidation by exposure to high-energy
X-ray irradiation (45-krad dose); D: RNase after
oxidation with 2.5 mM H2O2 and 10 jiM Cu 2 + , pH
7.4, for 4 h at room temperature. All samples
were adjusted to pH 2 for fluorescence analysis.
Protein concentrations were 1 mg/ml in A and
B and 5 mg/ml in C and D. Native RNase (not
shown), which lacks tryptophan, has negligible
visible wavelength fluorescence under these
conditions, i.e., 3-D graph is flat, whereas
collagen from nondiabetic or younger donors
has 3-D spectrum similar to that in A, except
for lower signal intensity.
Thus far, specific carbohydrate-derived products of au-
toxidative glycosylation, such as the ketoimine adduct to
protein, have not been identified in proteins either in vitro or
in vivo, so that the significance of this pathway remains con-
troversial (23,24). On the other hand, regardless of their
origins or mechanism of formation, by autoxidative gly-
cosylation or otherwise, CML, CMhL, and pentosidine are
sugar-derived autoxidation products, which have been iden-
tified in tissue proteins. They are formed by free radical ox-
idation reactions and may also participate in the initiation
and propagation of damaging free radical reactions. These
three compounds and total fluorescence increase coordi-
nately in collagen with age and at an accelerated rate in
diabetes (9-11,13), and thus they provide evidence of in-
creased oxidative damage to collagen in diabetes. Because
of the interplay between glycation and oxidation in their for-
mation, we have termed these compounds glycoxidation
products (21). Because they are not formed by reactions of
proteins with malondialdehyde or peroxidized lipids, gly-
coxidation products may be considered biomarkers of car-
bohydrate-dependent damage to protein and indicators of
the extent of underlying chemical modification, oxidation,
and cross-linking of tissue protein caused by reducing sug-
ars. Furthermore, because these products accumulate in
collagen normally as a function of age and at an accelerated
rate in diabetes, diabetes may be legitimately described, at
the chemical level, as a disease characterized by acceler-
ated aging of collagen by both glycative and oxidative mech-
anisms. Individual differences in the accumulation of gly-
coxidation products in collagen (2- to 3-fold ranges at ages
60-80 yr in both diabetic and nondiabetic populations) sug-
gest a wide variation in individual susceptibility to damage,
an observation that might yield insight into the basis for in-
dividual differences in susceptibility to development of com-
plications.
DIABETES, VOL. 40, APRIL 1991

IS GLYCOXIDATIVE DAMAGE SUFFICIENT TO CAUSE
OBSERVED STRUCTURAL CHANGES IN COLLAGEN?
The Amadori adduct FL is the major Maillard reaction product
identified in collagen but accounts for, at most, 2-3% of the
lysine residues in the protein or ~1 mol FL/mol triple-
stranded collagen, even in poorly controlled diabetes. There
is no evidence that this extent of modification is harmful, and
correlations between the level of glycation of collagen and
the presence of complications are weak (25,26), indicating
that glycation alone is not sufficient to cause complications.
In contrast, although glycoxidation products accumulate
gradually and irreversibly in collagen, consistent with a pos-
sible role in the development of complications, they are found
in only trace concentrations in the protein. For example, CML
is present at <10% of the level of FL, even in patients with
long duration of diabetes. This level of CML would have
minimal effects on the charge properties of collagen, and
because CML is not involved in cross-linking, it is not likely
to affect the protein's physical properties or resistance to
proteases (Fig. 1). Although the pentosidine content and
fluorescence of collagen are correlated with the presence
of complications (11,27), there are even lower levels of pen-
tosidine in collagen, —0.01 mol pentosidine/mol triple-
stranded collagen in the skin of elderly diabetic patients.
This is probably <1 % of the level of natural enzymatic cross-
links in skin collagen, and although up to 10-fold higher levels
of pentosidine are found in collagens in other tissues (10),
it is unlikely that pentosidine cross-links are sufficient to
cause significant changes in the properties of the protein.
Although pentosidine is present in only trace amounts in
collagen, it represents 25-40% of the fluorescence at 328/
378 nm in human skin collagen. Pentosidine also accounts
for - 8 0 % of the fluorescence in RNase or lysozyme browned
by incubation with glucose in vitro (250 mM glucose for 1
mo at 37°C in 0.2 M phosphate buffer), thus it is a major
fluorophore in both natural and artificially browned proteins.
Despite its prominence, however, the pentosidine content of
protein dimers isolated from glycation reaction mixtures ac-
counts for only a small fraction of the total intermolecular
cross-links in the dimerized protein (21), i.e., <0.01 mol pen-
tosidine/mol lysozyme dimer, whereas there must be at least
1 mol intermolecular cross-link/mol protein dimer. By an ag-
gressive extrapolation from experiments with model proteins
in vitro, we might conclude that pentosidine also accounts
for < 1 % of the carbohydrate-dependent nonenzymatic
cross-links present in collagen. Multiplying the pentosidine
content of collagen by 100 would yield a level of nonenzy-
matic cross-links in collagen approaching that of the enzy-
matic cross-links present in the protein. Thus, the trace con-
centration of pentosidine in collagen suggests an underlying
level of nonenzymatic cross-linking and perhaps other chem-
ical modifications, clearly sufficient to affect the physical and
chemical properties of the protein. Although their structures
are unknown, most of the nonenzymatic cross-links must be
relatively nonfluorescent, perhaps colorless compounds;
otherwise, they would be more apparent in the absorbance
and fluorescence spectrum of the protein. Furthermore, if
they are formed by oxygen radical reactions initiated by
reducing sugars or glucose adducts to protein, it is possible
that the carbon skeleton of glucose may not be directly in-
volved in their structure; i.e., the cross-links may be glucose-
DIABETES, VOL. 40, APRIL 1991
J.W. BAYNES
dependent but not glucose-derived structures. These are
not trivial considerations, because efforts aimed at charac-
terization of fluorescent products in browned proteins may
miss the major products involved in cross-linking of the pro-
tein, and similarly, efforts to label the cross-links with radio-
active glucose may yield only traces of radioactive products
in highly cross-linked proteins. In summary, although gly-
coxidation products (Fig. 2) and fluorescence (Fig. 3) pro-
vide insight into the relative level of glycative and oxidative
damage to protein, those products characterized thus far
are not present at concentrations high enough to have a
significant impact on the physical or chemical properties of
collagen (Fig. 1). However, the presence and accumulation
of these compounds in collagen suggest the presence of a
larger fraction of unidentified cross-links and suggest a cat-
alytic and structural role for glucose in the oxidation and
cross-linking of collagen in diabetes. Continued analysis of
glycated proteins and collagen will be necessary to char-
acterize the quantitatively more significant cross-links, to de-
termine their origin and concentration in proteins, and then
to assess their significance in diabetes.
LIPID PEROXIDATION IN DIABETES
Significant changes in lipid metabolism and structure also
occur in diabetes, particularly in patients with vascular com-
plications (28). In these cases, the structural changes are
clearly oxidative in nature, and oxidation of lipids in plasma
lipoproteins and in cellular membranes is associated with
the development of vascular disease in diabetes (recently
reviewed by Lyons [5]). However, epidemiological studies
suggest that the level of lipid peroxides in human plasma is
more associated with hypertriglyceridemia and vascular dis-
ease itself rather than directly with diabetes (29). In diabetic
rats, increased lipid peroxidation was also associated with
hypertriglyceridemia, but the oxidation and resultant toxicity
of the oxidized lipoproteins were inhibited by administration
of the lipophilic antioxidant probucol without an effect on the
hyperlipidemia (30). Probucol also inhibited the develop-
ment of atherosclerosis in nondiabetic hyperlipidemic rab-
bits (31), but its effectiveness in the treatment of diabetic
vascular disease has not been tested. These experiments
suggest that diabetes and hyperlipidemia are not alone suf-
ficient to induce vascular disease and argue that oxidative
stress may be an independent risk factor in the development
of vascular disease.
Research on the role of lipid peroxidation in diabetic com-
plications is hampered by the complexity of products formed
(32) and limitations in the various assays for measuring the
status and products of lipid peroxidation (33). Furthermore,
although there are numerous reports of increased peroxi-
dation of lipids in plasma lipoproteins, in erythrocyte mem-
brane proteins, and in various tissues in diabetes, the relative
importance of enzymatic versus nonenzymatic sources of
the lipid peroxidation in diabetes is unknown (3,5). Increased
lipid peroxides in plasma might result from enzymatic pro-
cesses activated by generalized vascular inflammation,
leading to increased levels of prostaglandins and lipoxy-
genase products. Alternatively, lipid peroxides might be
formed by nonenzymatic reactions of unsaturated lipids with
superoxide radical, hydrogen peroxide, and adventitious
metal ions in the circulation or extravascular space or at the
409
OXIDATIVE STRESS IN DIABETES
surface of endothelial and phagocytic cells. The distinction
between enzymatic and nonenzymatic (autoxidative) oxi-
dation of lipids in vivo is not absolute. Thus, the enzymatic
synthesis of prostaglandins may be stimulated by lipid per-
oxides derived from nonenzymatic pathways, and enzy-
matically generated lipid peroxides may also react with metal
ions to initiate autoxidation reactions. Hydrogen peroxide
and superoxide, intermediates in the autoxidative pathway,
are also produced by both enzymatic and nonenzymatic
pathways. Some lipid peroxidation products may be formed
by both pathways, and degradation products, measured by
the thiobarbituric acid assay, may also be derived from prod-
ucts of either pathway. Hayaishi and Shimizu (34) showed
that a significant decrease in total lipid peroxides in rabbit
plasma occurred within a few hours after aspirin administra-
tion. Similar experiments have not been conducted in hu-
mans, but the observation illustrates the probable involve-
ment of both enzymatic and nonenzymatic.pathways of lipid
peroxidation in diabetes and the sensitivity of lipid peroxi-
dation to anti-inflammatory agents. In general, studies on
lipid peroxidation are consistent with studies on glycoxida-
tion of proteins in diabetes; i.e., increased oxidation of both
lipids and proteins is associated with the development of
complications. However, comparative studies on oxidation
of lipids and proteins in diabetes have not been reported.
It is difficult to conclude whether increased lipid peroxi-
dation is a cause or effect of complications in diabetes (5),
and it is probably more appropriate to consider lipid per-
oxidation as part of a continuous cycle of oxidative stress
and damage. Lipid peroxidative damage may not be limited
to the lipid compartment because lipid peroxides may cause
browning and cross-linking of collagen (35) and contribute
to the development of fluorescence in plasma proteins
(and possibly collagen) in diabetes (20,36). This crossover
between the oxidative chemistry of lipids and proteins is
reminiscent of experiments discussed earlier in which gly-
cation of proteins causes oxidation of associated lipids
(14,15) and enhances the generation of fluorescence during
oxidation of proteins (18). Thus, increased glycation of col-
lagen and plasma proteins in diabetes may stimulate the
oxidation of lipids, which may in turn stimulate autoxidative
reactions of sugars, enhancing damage to both lipids and
proteins in the circulation and the vascular wall, continuing
and reinforcing the cycle of oxidative stress and damage.
In this case, it is less important to fix the blame and more
important to focus on the development of various possible
therapeutic approaches for intervening in the cyclic process.
OTHER EVIDENCE FOR ROLE OF OXIDATIVE STRESS IN
CROSS-LINKING OF COLLAGEN
There are indications in the literature that suggest that an-
tioxidant or anti-inflammatory therapy may limit damage to
proteins by glycation reactions. Thus, aspirin and salicylate
inhibited the increase in tail collagen cross-linking in diabetic
rats, as measured by effects on thermal rupture time (37).
This effect was observed without an effect on glycation, sug-
gesting that the drugs might be acting as inhibitors of oxi-
dation and oxidative cross-linking reactions through their
inhibition of cyclooxygenase activity. The drug dosages
used in these experiments were probably too high for ther-
410
apeutic use in humans, but similar effects were observed
with the lipoxygenase inhibitors indomethacin and naproxen
at doses within the therapeutic range (38), again without an
effect on glycation. The action of these drugs could result
either from inhibition of enzymatic pathways of lipid perox-
idation or by their action as oxygen radical scavengers. The
impressive therapeutic effect of sorbinil, an aldose reductase
inhibitor, on collagen-linked fluorescence (39) and vascular
permeability (40) in experimental animals could also be in-
terpreted as the result of their antioxidant activity, either di-
rectly by their action as oxygen radical scavengers or in-
directly by their effects on cellular redox potentials and
NADPH and glutathione concentrations (4,40). Rutin, an al-
dose reductase inhibitor and, based on its structure, prob-
ably a transition metal ion chelator and radical scavenger,
also inhibited the development of collagen-linked fluores-
cence in diabetic rats (41). Other effects of nonsteroidal anti-
inflammatory and antioxidant agents may be more general,
such as inhibition of neutrophil activation by inflammatory
stimuli (42), which would limit the systemic production of
free radicals and initiation or propagation of oxidative dam-
age by both carbohydrate-dependent and lipid-dependent
mechanisms.
Studies on the mechanism of action of aminoguanidine
(AG) also suggest that autoxidation reactions are involved
in the cross-linking of collagen by glucose. AG is the one
agent specifically designed to inhibit the browning and
cross-linking of protein by glucose during advanced stages
of the Maillard reaction (43). It works well in vitro and in vivo,
inhibiting the development of fluorescence, the formation of
pentosidine, and the cross-linking of collagen (43) and
model proteins (21) by glucose but is without an effect on
glycation of the proteins (41,43,44). AG is not an antioxidant,
based on the fact that it does not inhibit the formation of
superoxide from Amadori compounds or the oxidation of FL
to CML in vitro. However, we have observed that oxygen
accelerates the cross-linking of collagen by glucose in vitro,
indicating that autoxidation reactions may be important in
the formation of fluorescent and cross-linking products in
collagen in vivo. During its inhibition of cross-linking, AG
forms characteristic carbohydrate adducts in solution, and
these compounds appear to be derived from its reaction
with products of oxidation of glucose. These products are
formed at similar rates in the presence or absence of col-
lagen, suggesting that AG is trapping dicarbonyl interme-
diates formed by autoxidation of glucose. Further studies on
the mechanism of action of AG should clarify the role of
autoxidation reactions in the cross-linking of proteins by glu-
cose in vitro and also suggest methods for evaluating the
role of sugar autoxidation in vivo.
CONCLUSIONS
To the landscape artist, "perspective" deals with the con-
vergence of lines to portray relationships between objects.
This article has dealt with convergence and relationships,
presenting the hypothesis that oxidative stress may be a
common pathway relating diverse seemingly distinct mech-
anisms proposed for the pathogenesis of complications in
diabetes. I have tried to argue not that oxidative stress is
increased in diabetes, which then leads to the development
DIABETES, VOL. 40, APRIL 1991

of complications, but primarily that diabetes with compli-
cations is associated with increased chemical modification
of proteins and lipids and that this "damage" appears to be
largely oxidative in origin and is sufficient to explain the
altered function of proteins in the extracellular matrix. There
are many possible causes of increased oxidative stress in
diabetes, but the source of the oxidative stress, if there is
one primary source, may be extremely difficult to determine,
especially if cyclic, autocatalytic, and reinforcing processes
are involved. At this stage, fundamental information is
needed on the nature of products formed during oxidation
of proteins and of products accumulating in collagen in di-
abetes, so that the relationship between oxidative stress and
the development of complications in diabetes can be ad-
dressed more directly. It would also be worthwhile to identify
discrete products whose level in blood proteins could be
used as short-term or medium-term integrators of oxidative
stress in the manner in which glycation of plasma proteins
and hemoglobin is used as an index of glycemic stress.
These products, distinct from long-term integrators such as
glycoxidation products in collagen, would provide an indi-
cation of the current status of oxidative stress rather than
cumulative oxidative damage. They may also be useful for
identifying patients at risk or with incipient disease and for
assessing responses to antioxidant therapy. Eventually,
these studies may lead to the development of effective strat-
egies for limiting the damage from glycation and oxidation
of proteins or for complementing other therapeutic ap-
proaches to the treatment of complications in diabetes.
ACKNOWLEDGMENTS
Research in my laboratory was supported by Research Grant
DK-19971 from the National Institute of Diabetes and Diges-
tive and Kidney Diseases.
I thank Dr. Suzanne R. Thorpe and Dr. Timothy J. Lyons
for helpful discussions and Dr. Daniel G. Dyer, Dr. Thomas
G. Huggins, and James A. Blackledge for the experimental
data shown in Fig. 3.
REFERENCES
1. Halliwell B, Gutteridge JMC: Free Radicals in Biology and Medicine. 2nd
ed. Oxford, UK, Oxford Univ. Press, 1989
2. Oberly LW: Free radicals and1 diabetes. Free Radical Biol Med 5:113-24,
1988
3. Godin GV, Wohaieb SA: Reactive oxygen radical processes in diabetes.
In Oxygen Radicals in the Pathophysiology of Heart Disease. Singal PK,
Ed. Boston, MA, Kluwer, 1988
4. Wolff SP: The potential role of oxidative stress in diabetes and its com-
plications: novel implications for theory and therapy. In Diabetic Com-
plications: Scientific and Clinical Aspects. Crabbe MJC, Ed. New York,
Churchill-Livingstone, 1987, p. 167-220
5. Lyons TJ: Oxidised low density lipoproteins—a role in the pathogenesis
of atherosclerosis in diabetes. Diabetic Med. In press
6. Daugherty A, Baynes JW (Eds.): The Role of Oxidation in Pathophysiol-
ogy. St. Louis, MO, MedStrategy, 1991
7. Stadtman ER: Metal ion-catalyzed oxidation of proteins: biochemical
mechanism and biological consequences. Free Radical Biol Med 9:315-
25, 1990
8. Ahmed MU, Thorpe SR, Baynes JW: Identification of N'-(carboxymethyl)-
lysine as a degradation product of fructoselysine in glycated protein. J
Biol Chem 261:4889-94, 1986
9. Dunn JA, McCance DR, Thorpe SR, Lyons TJ, Baynes JW: Age-depen-
dent accumulation of N€-(carboxymethyl)lysine and N£-(carboxy-
methyl)hydroxylysine in human skin collagen. Biochemistry. In press
10. Sell DR, Monnier VM: Structure elucidation of a senescence cross-link
from human extracellular matrix: implication of pentoses in the aging
process. J Biol Chem 264:21597-602, 1989
DIABETES, VOL. 40, APRIL 1991
J.W. BAYNES
11. Sell DR, Monnier VM: End-stage renal disease and diabetes catalyze the
formation of a pentose-derived cross link from aging human collagen. J
Biol Chem 85:380-84, 1990
12. Ahmed MU, Dunn JA, Walla MD, Thorpe SR, Baynes JW: Oxidative deg-
radation of glucose adducts to protein: formation of 3-(Ne-lysino)-lactic
acid from model compounds and glycated proteins. J Biol Chem
263:8816-21, 1988
13. Kaanane A, Labuza TP: The Maillard reaction in food. In The Maillard
Reaction in Aging, Diabetes and Nutrition. Baynes JW, Monnier VM, Eds.
New York, Liss, 1989
14. Hunt JV, Smith CCT, Wolff SP: Autoxidative glycosylation and possible
involvement of peroxides and free radicals in LDL modification by glu-
cose. Diabetes 39:1420-24, 1990
15. Hicks M, Delbridge L, Yue DK, Reeve TS: Catalysis of lipid peroxidation
by glucose and glycosylated proteins. Biochem Biophys Res Commun
151:649-55, 1988
16. Jones AF, Winkles JW, Thornalley PJ, Lunec J, Jennings PE, Barnett AH:
Inhibitory effect of superoxide dismutase on fructosamine assay. Clin
Chem 33:147-49, 1987
17. Sakurai T, Tsuchiya S: Superoxide production from nonenzymatically gly-
cated protein. FEBS Lett 236:406-10, 1988
18. Fujimori E: Cross-linking and fluorescence changes of collagen by gly-
cation and oxidation. Biochim Biophys Acta 998:105-10, 1989
19. Wickens DG, Norden AG, Lunec J, Dormandy TL: Fluorescence changes
in human gamma globulin induced by free radical activity. Biochim Bio-
phys Acta 742:607-16, 1983
20. Jones AF, Lunec J: Protein fluorescence and its relationship to free radical
activity. BrJ Cancer (Suppl. VIII) 56:60-65, 1987
21. Dyer DG, Blackledge JA, Katz BM, Hull CJ, Adkisson HD, Thorpe SR,
Lyons TJ, Baynes JW: The Maillard reaction in vivo. J Nutr Sci 29:13-20,
1990
22. Jiang ZY, Woollard ACS, Wolff SP: Hydrogen peroxide production during
experimental glycation. FEBS Lett 268:69-71, 1990
23. Harding JJ, Beswick HT: The possible contribution of glucose autoxi-
dation to protein modification in diabetes. Biochem J 249:617-18, 1988
24. Wolff SP, Dean RT: Aldehydes and dicarbonyls in non-enzymic glyco-
sylation of proteins. Biochem J 249:618-19, 1988
25. Vishwanath V, Frank KE, Elmets CA, Dauchot PJ, Monnier VM: Glycation
of skin collagen in type I diabetes mellitus: correlation with long-term
complications. Diabetes 35:916-21, 1986
26. Lyons TJ, Kennedy L: Nonenzymatic glycosylation of skin collagen in
patients with type I (insulin-dependent) diabetes mellitus and limited joint
mobility. Diabetologia 28:2-5, 1985
27. Monnier VM, Vishwanath BA, Frank KE, Elmets CA, Dauchot P, Kohn RR:
Relation between complications of type I diabetes mellitus and collagen-
linked fluorescence. N Engl J Med 314:403-408, 1986
28. Sato Y, Hotta N, Sakamoto N, Matsuoka S, Ohishi N, Yagi K: Lipid peroxide
level in plasma of diabetic patients. Biochem Med 25:373-78, 1981
29. Stringer MD, Gdrbg PG, Freeman A, Kakkar W : Lipid peroxides and
atherosclerosis. Br Med J 298:281-84, 1989
30. Morel DW, Chisolm GM: Antioxidant treatment of diabetic rats inhibits
lipoprotein oxidation and cytotoxicity. J Lipid Res 30:1827-34, 1989
31. Kita T, Nagano Y, Yokode M, Ishii K, Kume N, Ooshima A, Yoshida H,
Kawai C: Probucol prevents the progression of atherosclerosis in Watan-
abe heritable hyperlipidemic rabbit, an animal model for familial hyper-
cholesterolemia. Proc Natl Acad Sci USA 84:5928-31, 1987
32. Esterbauer H, Jurgens G, Quehenberger Q, Koller E: Autoxidation of
human low density lipoprotein: loss of polyunsaturated fatty acids and
vitamin E and generation of aldehydes. J Lipid Res 28:495-509, 1987
33. Gutteridge JMC, Halliwell B: The measurement and mechanism of lipid
peroxidation in biological systems. Trends Biochem Sci 15:129-35,1990
34. Hayaishi O, Shimizu T: Metabolic and functional significance of prosta-
glandins in lipid peroxide research. In Lipid Peroxides in Biology and
Medicine. Yagi K, Ed. New York, Academic, 1982
35. Pokorny J, Davidek J, Chocholata V, Panek J, Bulantova H, Janitz W,
Valentova H, Vieredklova M: Interactions of oxidized lipids with protein.
Nahrung 34:159-69, 1990
36. Tsuchida M, Miura T, Mizutani K, Aibara K: Fluorescent substances in
mouse and human sera as a parameter of in vivo lipid peroxidation.
Biochim Biophys Acta 834:196-204, 1985
37. Yue DK, McLennan S, Handelsman DJ, Delbridge L, Reeve T, Turtle JR:
The effect of salicylates on nonenzymatic glycosylation and thermal sta-
bility of collagen in diabetic rats. Diabetes 33:745-51, 1984
38. Yue DK, McLennan S, Handelsman DJ, Delbridge L, Reeve T, Turtle JR:
The effects of cyclooxygenase and lipoxygenase inhibitors on the col-
lagen abnormalities of diabetic rats. Diabetes 34:74-78, 1985
39. Suarez G, Rajaram R, Bhuyan KC, Oronsky AL, Gold JA: Administration
of an aldose reductase inhibitor induces a decrease of collagen fluores-
cence in diabetic rats. J Clin Invest 82:624-27, 1988
40. Williamson JR, Ostrow E, Eades D, Chang K, Allison W, Kilo C, Sherman
WR: Glucose-induced microvascular functional changes in nondiabetic
rats are sterospecific and are prevented by an aldose reductase inhibitor.
411
OXIDATIVE STRESS IN DIABETES
J Clin Invest 85:1167-72, 1990
41. Odetti PR, Borgoglio A, De Pascale A, Rolandi R, Adezati L: Prevention
of diabetes-increased aging effect on rat collagen-linked fluorescence
by aminoguanidine and rutin. Diabetes 39:796-801, 1990
42. Abramson AB, Weissmann G: The mechanisms of action of non-steroidal
anti-inflammatory drugs. Arthritis Rheum 32:1-9, 1989
412
43. Brownlee M, Vlassara H, Kooney A, Ulrich P, Cerami A: Aminoguanidine
prevents diabetes-induced arterial wall protein cross-linking. Science
232:1629-32, 1986
44. Nicholls K, Mandel TE: Advanced glycosylation end-products in experi-
mental murine diabetic nephropathy: effect of islet isografting and of
aminoguanidine. Lab Invest 60:486-91, 1989
DIABETES, VOL. 40, APRIL 1991