ARCHIVES OF BIOCHEMISTRY AND BIOPHYSICS Vol. 268, No. 1, January, pp. 249-254,1989

Increase in Crosslinking of Nonenzymatically Glycosylated Collagen

Induced by Products of Lipid Peroxidation

MARK HICK&l LEIGH DELBRIDGE, DENNIS K. YUE, AND TOM S. REEVE

Department of Surgery, The University of Sydney, Royal North Shore Hospital, St. Leonads, New South Wales 2065, Australia

Received May 9,1988, and in revised form September 9,1988

The increase in crosslinking in normal and nonenzymatically glycosylated rat tail ten- don collagen after treatment with decomposing lipid hydroperoxides was assessed by measuring the breaking time of tendons immersed in 7 M urea under a 3 g weight at 40°C (thermal rupture time). The incubation of tendons in 200 mM glucose for 43 h at 40°C increased thermal rupture times from 5.15 to 26.38 min, (P < 0.001) with no significant corresponding increase in tendons incubated in buffer alone. After incubation of the glycosylated tendons in the presence of peroxidized linoleic/arachidonic acid vesicles for about 20 h, their thermal rupture time increased to 3360 min (P < 0.001). The rupture time for normal tendons after the same treatment was 206 min. These apparent cross- linking increases cannot be fully accounted for by reactions involving malondialdehyde, as incubation of both glycosylated and normal tendons in enzymatically produced ma- londialdehyde resulted in a modest two- to threefold increase in thermal rupture time. 0 1989 Academic PRSS, 112~.

Crosslinking of collagen peptides occurs as a normal part of the maturation of this structural protein (1). During diabetes mellitus, crosslinking is increased, and is thought to be mediated by attachment of glucose to basic amino acid residues during hyperglycemia. This nonenzymatic glyco- sylation is thought to be followed by a fur- ther series of chemical rearrangements which result in highly fluorescent glucose- derived crosslinks (termed advanced gly- cosylation endproducts) (2). Such cross- linking has been postulated to play an im- portant role in many of the complications associated with the diabetic state (3).

Recent evidence has raised some doubts as to the direct role of the glycosylated res- idues in the crosslinking process for a number of reasons. (i) Nonsteroidal anti- inflammatory agents such as indometha- tin which, in contrast to aspirin do not

’ To whom correspondence should be addressed

chemically alter the lysine residue, are also able to inhibit crosslinking in the dia- betic rat (4). (ii) Compounds arising from oxidative cleavage of the initial glycosyl- ated residues as well as reversal of glyco- sylation have been observed (5). (iii) Fluo- rescence increases which do not correlate with the level of nonenzymatic glycosyl- ation have been found in proteins with short half-lives such as IgG and albu- min from diabetic subjects with retinopa- thy (6).

These observations raise the possibility that molecules other than sugars may be involved in the crosslinking process. It is well known that crosslinking of proteins, lipids, and DNA occurs in the presence of a range of products associated with lipid hydroperoxide decomposition (7,8). In dia- betic patients, especially those with vascu- lar complications, parameters associated with lipid peroxidation such as diene con- jugation (9), Schiff’s base fluorescence (lo), and malondialdehyde (11,12) are markedly

249 0003-9861/89 $3.00 Copyright 0 1389 by Academic Press. Inc. All rights of reproduction in any form reserved.

250 HICKS ET AL.

increased in comparison with normogly- cemic controls. In addition we have shown that peroxidation of lipids in vitro can be initiated by free glucose or glycosylated collagen (13). To date, the potential role of lipid hydroperoxide decomposition prod- ucts in crosslinking proteins in diabetes has not been assessed.

The aim of the present study was to in- vestigate the relationship between in vitro nonenzymatic glycosylation of rat tail ten- don collagen and crosslinking induced by hydroperoxide decomposition products from peroxidized polyunsaturated fatty acid vesicles.

MATERIALS AND METHODS

Linoleic and arachidonic acids were obtained from Sedary Research Laboratories (London, Ontario), tris(hydroxymethyl)aminomethane from Sigma Chemical Co. (St. Louis, MO) and 1,3-propanediol from Aldrich Chemical Co. (Milwaukee, WI). Lactate dehydrogenase, alcohol dehydrogenase, NAD, and py- ruvate were purchased from Boehringer-Mannheim (Australia) and methanol (GR grade) from Merck (Darmstadt). All other chemicals were of AR grade and were bought from BDH (Australia). Water used for solution preparation and glassware washing was passed through a Mini-Q water purification unit (Millipore, Pty. Ltd., Australia).

Rat tail tendons were used as the source of collagen. Male Wistar rats (8-10 weeks old) were obtained from the Atomic Energy Commission (Lucas Heights). Af- ter sacrifice, tail tendons were removed, washed thor- oughly in physiological saline and distilled water, then dried and stored at -70°C until required.

Nonenzymatic glycosylation of tendon collagen was carried out as follows. Tendon fibres (10 cm in length, weighing 1-2 mg) were incubated at 40°C for 43 h in 200 mM glucose in 0.066 M disodium hydrogen/potas- sium dihydrogen orthophosphate, at pH 7.4, which contained 0.2% azide as an antibacterial agent. The extent of glycosylation was measured by thiobarbi- turic acid chromophore absorbance at 443 nm (14), af- ter solubilization of the tendon with collagenase (15). Hydroxymethyl furfural (HMF)’ was used as the standard. Collagen was determined by the method of Stegemann and Stalder (16). The relative extent of glycosylation was expressed as nmol HMF/mg col- lagen.

The relative extent of tendon collagen crosslinking was assessed by measuring the thermal rupture time

’ Abbreviations used: HMF, hydroxymethyl furfu- ral; AGES, advanced glycosylation endproducts.

of the fibers in urea (17). Each lo-cm fiber to be tested was cut in half, then suspended between the arm of an electronic timing device and a 3-g weight. The ten- don was immersed in 7 M urea at 40°C. The breaking time was recorded by an electronic timing apparatus. The mean breaking time for four tendons was taken as the thermal rupture time.

Fatty acid vesicles were prepared by the method of Gebicki and Hicks (18). A mixture of linoleic acid (40 mg) and arachidonic acid (0.4 mg) was mixed with 4 ml 0.1 M Tris/HCl, pH 8, in a test tube, then the sus- pension was vortexed for 2 min. The resulting vesicle preparation was then diluted to 20 ml with 0.066 M disodium hydrogen/potassium dihydrogen ortho- phosphate buffer, pH 8 (final concentration of fatty acid 7.2 mM). Peroxidation of fatty acids in the vesicle membranes was induced by incubating the prepara- tions at 50°C for 3 days. Hydroperoxide content was measured by absorbance at 234 nm due to conjugated diene formation, using an extinction coefficient of 28,000 liters. mall’ cm-’ (19). Malondialdehyde for- mation was measured by the method of Ohkawa et al (20).

Malondialdehyde was also prepared in the absence of lipid by a modification of the method of Summer- field and Tappel(21). The reaction mixture contained 5 ml 100 pM NAD, 2 ml 100 mM pyruvate, 2 ml 5 mM 1,3-propanediol, all in 0.066 M disodium hydrogen/po- tassium dihydrogen phosphate buffer, pH 9, and 1 ml phosphate buffer. The reaction was started by addi- tion of 1.35 x 10m4 units of alcohol dehydrogenase and 2 X lo-* units of lactate dehydrogenase.

RESULTS

The effect of in vitro glycosylation of ten- don collagen on thermal rupture time is shown in Fig. 1. Incubation of tendons in 200 mM glucose resulted in an approximate fivefold increase in mean thermal rupture time over the control tendons incubated in phosphate buffer alone after 43 h, (5.15 + 1.3 to 26.38 f 2.2, n = 8, P -c 0.001). The extent of glycosylation had increased from 3.8 to 10 nmol HMF/mg collagen in the glu- cose treatment during this period, while the extent of glycosylation in the buffer treatment remained unchanged at 3.8 nmol HMF/mg collagen.

After 43 h, batches of tendons were re- moved from both glucose and buffer treat- ments, washed thoroughly in distilled wa- ter, then transferred to aliquots of peroxi- dized polyunsaturated fatty acid vesicle preparation or buffer alone, and incubated at 40°C. Tendons were removed at approxi-

CROSSLINKING IN GLYCOSYLATED COLLAGEN 251

TIME (h)

FIG. 1. Increase in thermal rupture times of tendon collagen as a result of in vitro nonenzymatic glycosyl- ation. All points were the means of the determina- tions of eight tendons with error bars representing standard deviations. Tendons incubated in 200 mM glucose (0); tendons incubated in phosphate buffer (0). Both treatments contained 0.2% azide. Incuba- tion temperature, 40°C.

mately 5-h intervals for measurement of mechanical breaking strength.

Thermal rupture times for both nonen- zymatically glycosylated and control ten- dons increased significantly over the time course of the experiment (Fig. 2). For ten- dons that had been preincubated in glu- cose, mean thermal rupture time increased from 26.38 f 2.2 to 3360 + 100 min (about 130-fold). For tendons that had been prein- cubated in buffer, thermal rupture time in- creased from 5.15 -t 1.3 to 206 + 10 min (3% fold). The thermal rupture times for glyco- sylated and buffer-treated tendons after 22.5-h incubation in peroxidized lipid were significantly different at the P < 0.001 level (Fig. 2). There was no significant difference between the thermal rupture times of ten- dons that had no prior treatment before incubation in peroxidized lipid and those preincubated in buffer (data not shown). Thus the presence of azide in the preincu- bation buffers (itself a weak oxidizing agent) could not account for any artifac- tual increases in collagen crosslinking.

Changes in hydroperoxide and free ma- londialdehyde content of peroxidized fatty acid vesicle preparations during tendon in- cubation are shown in Figs. 3a and 3b, re-

spectively. At the time of addition of ten- dons, hydroperoxides were already decom- posing and the rate of this decomposition was unaffected by the presence of tendons (Fig. 3a). The level of free malondialde- hyde in the vesicle preparation also fell during the time course of the experiment. The rate of malondialdehyde loss was greater in the presence of tendons than in their absence. Once again the rate of ma- londialdehyde decrease was independent of the nature of the tendons (i.e., nonenzy- matically glycosylated, buffer treated, or untreated).

Decomposing lipid hydroperoxides con- tain a number of aldehyde and hydroxy-al- kenal derivatives which are capable of initiating crosslinking in proteins (7, 8). Malondialdehyde has been a popular can- didate for induction of such crosslinking. In order to test the role of this bifunctional aldehyde in the observed increases in ther- mal rupture time in glycosylated tendon collagen, tendons from the experiment de- scribed in Fig. 1 were incubated in the re- action mixture for enzymatic malondialde- hyde production at 20°C. Figure 4 shows

TIME (hl

FIG. 2. Increase in thermal rupture times of gly- cosylated and control tendon collagen on exposure to peroxidized linoleic/arachadonic acid vesicles. Twenty tendons from each treatment were added at t = 0. All points were the means of determinations of eight tendons with error bars representing standard deviations. Where bars are not shown, the standard deviations lay within the diameter of the points. Ten- dons previously nonenzymatically glycosylated (0); tendons pretreated in buffer (0). Incubation tempera- ture, 40°C.

252 HICKS ET AL.

TIME (h) TIME (h)

FIG. 3. (a) Kinetics of hydroperoxide loss in linoleic/arachidonic acid vesicles after addition of tendons (at t = 0). Hydroperoxide decrease in the presence of glycosylated (0); buffer-treated (0); untreated (A); no (0) tendon. (b) Kinetics of malondialdehyde loss in linoleic/arachidonic acid vesi- cles after addition of tendons (at t = 0). Malondialdehyde loss in the presence of glycosylated (D); buffer-treated (0); untreated (A); no (+) tendon.

the increases in thermal rupture times for glycosylated and buffer-treated tendons. Breaking times for both tendon treat- ments increased at a similar rate over the time course of the experiment (2.3- and 2.7- fold increases for glycosylated and control treatments, respectively).

Free malondialdehyde levels during this experiment are shown in Fig. 5. Amounts of malondialdehyde produced over the time course of the experiment are of the same order of magnitude as those present in the decomposing lipid hydroperoxide

FIG. 4. Thermal rupture time for tendons exposed to enzymatically produced malondialdehyde. Glyco- sylated (0); untreated (0) tendon. Twenty tendons from each treatment were added at t = 0. Each point was the mean of determinations of four tendons with error bars representing standard deviations. Incuba- tion temperature, 20%.

system shown in Fig. 3b. The rate of ma- londialdehyde production in the sample containing glycosylated tendons was higher than that for either the sample with normal or no added tendons. How- ever, this was not reflected in any increase in the rate of thermal rupture times.

DISCUSSION

Many biochemical and physical changes have been demonstrated in collagen during the onset of diabetes (15, 22-24). Rises in thermal rupture time are indicative of an increase in the extent of covalent cross-

TIME (h)

FIG. 5. Kinetics of enzymatically produced malon- dialdehyde. See Materials and Methods for composi- tion of reaction mixture. Malondialdehyde generat- ing system in the presence of glycosylated (0); un- treated (0); or no (0) tendons.

CROSSLINKING IN GLYCOSYLATED COLLAGEN 253

linking (15,25). This crosslinking has been attributed to a series of slow chemical re- arrangements subsequent to nonenzy- matic glycosylation to form “advanced gly- cosylation endproducts” (AGES) which contain highly fluorescent glucose-derived crosslinks (2,26).

In vivo, however, other potential cross- linking mechanisms may be possible in the diabetic state. Increased levels of products of lipid peroxidation have been detected in the serum of diabetic patients (9-12). De- composition products of hydroperoxides have been shown to be efficient crosslink- ing agents, giving products with fluores- cence characteristics similar to those of AGES (8, 27). Free radical induced oxida- tive damage has also been proposed to ac- count for fluorescent changes in short half- life proteins associated with diabetic mi- croangiopathy (28).

The present study demonstrates a sig- nificant increase in the extent of crosslink- ing (as determined by increases in thermal rupture time) of in vitro nonenzymatically glycosylated collagen when it is exposed to decomposing lipid hydroperoxides (Fig. 2). This increase is over and above that pro- duced by decomposing lipid hydroperox- ides on normal tendons (130-fold for glyco- sylated tendons compared to a 38-fold in- crease for normal tendons). During this crosslinking process, the levels of hydro- peroxides and free malondialdehyde in the vesicle preparation decreased. However, it was found that malondialdehyde produced in the absence of peroxidized lipid by an enzymatic process, caused only a modest increase in crosslinking (a 2- to 3-fold in- crease in thermal rupture time for both glycosylated and normal tendon collagen) and is at best only partially responsible for the observed thermal rupture increases in the presence of decomposing lipids perox- ides.

Enhanced crosslinking in collagen gels in the presence of oxidizing lipids has been demonstrated previously but the agents responsible for these increases were not identified (29). Recent studies summarized by Kikugawa and Beppu (8) have impli- cated a range of monofunctional aldehydes arising from lipid hydroperoxide decompo-

sition, including 2,4-hexadienal, 2,3-butad- ione, and 4-hydroxynonenal in formation of crosslinked, fluorescent protein deriva- tives. However, the potential role of these products on crosslinking of glycosylated proteins in the present system was not ad- dressed.

The results presented here, along with previous evidence which demonstrates the initiation of lipid peroxidation by free glu- cose or glycosylated species (13) suggest a mechanism for some of the degenerative changes to blood vessels which result from aging or diabetes mellitus. The long-lived collagen in arterial walls is glycosylated at a slow rate during the aging process and at a rapid rate during the hyperglycemic phase of diabetes (30). Hyperglycemia also would exert an oxidative stress on endo- thelial membrane lipids. Water-soluble decomposition products of hydroperoxides could then diffuse into the adjacent con- nective tissue layer and induce the colla- gen crosslinking and vessel wall hardening which precede the onset of more serious vascular complications.

ACKNOWLEDGMENTS

This work was supported by grants from CSL-Novo Pty., Ltd., the Kellion Diabetes Foundation, and the Surgical Trust Fund of the Royal North Shore Hospi- tal.

REFERENCES

1. BAILEY, A. J. (1978) J. Clin Pathol. 31, S~ppl. (R. Coil. Pa,th,ol.) 12,49-58.

2. BROWNLEE, M., VLASSARA, H., AND CERAMI, A. (1984) Ann. Int. Med 101,527-537.

3. KENNEDY, L., AND BAYNES, J. W. (1984) Diabeto- logia 26,93-98.

4. YUE, D. K., MCLENNAN, S., HANDELSMAN, D. J., DELBRIDGE, L., REEVE, T. S., AND TURTLE, J. R. (1985) Diabetes 34,74-X3.

5. AHMED, M. U., THORPE, S. R., AND BAYNES, J. W. (1986) J. BioL Chem. 261,4889-4894.

6. JONES, A. F., JENNINGS, P. E., WAKEFIELD, A., WINKLES, J. W., LUNEC, J., AND BARNETT, A. H. (1986) Lancet 2,459.

‘7. DILLARD, C. J., AND TAPPEL, A. L. (1984) in Meth- ods in Enzymology (Packer, L., Ed.), Vol. 105, pp. 337-341, Academic Press, San Diego.

8. KIKUGAWA, K., AND BEPPU, M. (1987) C&em. Phgs. Lipids 44,277-296.

254 HICKS ET AL.

9. JENNINGS, P.E., JONES,A.F.,FLORKOWSKI,C. M., LuNEc,J., AND BARNETT, A. H.(1987) Diabetic Med.4,452-456.

10. TSUCHIDA,M.,MIURA,T.,MIZUTANI,K.,ANDAIB- ARA, K. (1985) Biochim. Biophys. Acta834,196- 204.

11. SATO, Y., SAKAMOTO, N.,MATSUOKA, S., OHISHI, N., AND YAM, K. (1979) Biochem. Med. 21,104- 107.

12. NISHIGAKI, I., HAGIHARA, M., TSUNEKAWA, H., MASEKI,M.,ANDYAGI,K.(~~~~) Biochem.Med. 25,373-378.

13. HICKS,M.,DELBRIDGE,L.,YUE,D.K.,ANDREEVE, T. S. (1988) Biochem. Biophys. Res. Commun. 1X,649-655.

14. PECORARO, R. E., GRAF, R. J., HALTER, J. B., BEITER, H., AND PORTE, D.(1979)Diabetes 28, 1120-1125.

15. KOHN, R. R., CERAMI, A., AND MONNIER, V. M. (1984) Diabetes 33,57-59.

16. STEGEMANN, H., AND STALDER, K. (1967) C&n. Chim. Acta 18.267-273.

17. BOROS-FARKAS, M., AND EVERITT, A. V. (1967) Gerontologia 13,37-49.

18. GEBICKI,J.M.,ANDHICKS,M.(~~~~) Chem. Phys. Lipids 16,142-160.

19. PITT, G. A. J., AND MORTON, R. A. (1957) Prog. Chem. Fats Liyid.s 4,228-278.

20. OHKAWA, H., OHISHI, N., AND YAGI, K.(1978)J. LipidRes 19,1053-1057.

21. SUMMERFIELD, F. W., AND TAPPEL, A. L. (1978) Biochem. Biophys. Res. Commun. 82,547-552.

22. YUE, D. K., MCLENNAN, S., DELBRIDGE,L., HAN- DELSMAN,D.J.,REEVE,T., AND TURTLE,J. R. (1983) Diabetologia 24,282-285.

23. ANDREASSEN,T.T.,ANDOXLUND,H.(~~~~)D~~~~- tologia 28,687-691.

24. ANDREASSEN,T.T.,OXLUND,H.,ANDDANIELSEN, C. C.(1988) Connect. Tissue Res. 17,1-9.

25, RIGBY, B. J., MITCHELL, T. W., AND ROBINSON, M. S. (1977) Biochem. Biophys. Res. Commun. 79,400-405.

26. PONGER,~., ULRICH,~. C.,BENCSATH,F. A., AND CERAMI, A. (1984) Proc. NatL Acad. Sci USA 81,2684-2688.

27. JONES, A.F., ANDLUNEC,J. (1987) Brit. J. Cancer 55,Suppl. VIII,60-65.

28. JONES, A. F., JENNINGS, P. E., WAKEFIELD, A., WINKLES,J., LuNEc,J., AND BARNETT, A. H. (1986) New EngL J Med. 315,323-324.

29. SUNDHOLM, F., VISAPAA, A., AND BJORKSTEN,J. (1978) Lipids 13,755-757.

30. CERAMI, A., VLASSARA, H., AND BROWNLEE, M. (1987) Sci. Amer. 256,82-88.