Zinc-α2-Glycoprotein Is An Inhibitor Of Amine Oxidase Copper-Containing 3

Zinc-alpha2-glycoprotein (ZAG) is a major plasma protein whose levels increase in chronic energy-demanding diseases and thus serves as an important clinical biomarker in the diagnosis and prognosis of the development of cachexia. Current knowledge suggests that ZAG mediates progressive weight loss through β-adrenergic signaling in adipocytes, resulting in the activation of lipolysis and fat mobilization. Here, through crosslinking experiments, amine oxidase copper-containing 3 (AOC3) is identified as a novel ZAG binding partner. AOC3 – also known as vascular adhesion protein 1 (VAP-1) and semicarbazide sensitive amine oxidase (SSAO) – deaminates primary amines, thereby generating the corresponding aldehyde, H2O2 and HN3. It is an ectoenzyme largely expressed by adipocytes and induced in endothelial cells during inflammation. Extravasation of immune cells depends on amine oxidase activity and AOC3-derived H2O2 has an insulinogenic effect. The observations described here suggest that ZAG acts as an allosteric inhibitor of AOC3 and interferes with the associated pro-inflammatory and anti-lipolytic functions. Thus, inhibition of the deamination of lipolytic hormone octopamine by AOC3 represents a novel mechanism by which ZAG might stimulate lipolysis. Furthermore, experiments involving overexpression of recombinant ZAG reveal that its glycosylation is co-regulated by oxygen availability and that the pattern of glycosylation affects its inhibitory potential. The newly identified protein interaction between AOC3 and ZAG highlights a previously unknown functional relationship, which may be relevant to inflammation, energy metabolism and the development of cachexia.


28
Zinc-α2-glycoprotein (ZAG) was first isolated from human plasma more than 50 years ago. Its 29 name derives from its physicochemical properties, as it precipitates with bivalent ions such as fraction. The gel was stained with Coomassie Brilliant Blue and bands excised with a scalpel 133 (Fig. 1, Ad). For orientation, a WB of non-reduced samples probed with streptavidin was 134 carried out in parallel. Excised bands were prepared and subjected to mass spectrometry-135 based peptide sequencing. One excised band contained ZAG and identified SSAO (Fig. 1, B, 136 red box)from this point named AOC3as a putative interaction partner. AOC3 has a 137 molecular weight of ~84 kDa and exists as a homodimer. Given this, the shift of the GST 138 signal to a higher molecular weight under non-reducing conditions (Fig. 1, Ac) can be 139 explained by binding between one homodimeric AOC3 and at least one GST-ZAG molecule. 140 To confirm the newly identified protein interaction, it was attempted to purify AOC3 from E. 141 coli. Since all expression conditions failed, a modified method using HEK293 cells as 142 expression host was chosen [48]. Using lentivirus, secretable forms of GST-AOC3 ( 143 Fig. 10) and GST-tag were stably expressed in HEK293 cells. Both proteins were affinity 144 purified from the conditioned medium. To ascertain whether recombinant GST-AOC3 interacts 145 with murine plasma ZAG, a GST-pulldown was performed (Fig. 1, C). Plasma of overnight-146 7 fasted C57Bl6 male wt mice was incubated with recombinant GST-AOC3 and GST as a 147 control. A WB of the eluate fraction revealed that GST-AOC3 bound ZAG from murine 148 plasma, whereas GST alone did not. whether ZAG can modulate AOC3 activity, both GST-tagged AOC3 and ZAG of murine origin 162 were purified from lentivirally transduced HEK293 cells and the GST-tag was removed. In all 163 control assays, ZAG was replaced by the same amount of GST purified from stably transfected 164 HEK293 cells. In a first attempt, activity assays were performed using Amplex Red reagent 165 (Fig. 2, A). A saturation curve using benzylamine as substrate revealed the highest activity at 166 100 µM (Fig. 2, B). To characterize the interaction between AOC3 and ZAG, both proteins were 167 mixed at different molar ratios. A stepwise increase in the concentration of recombinant ZAG 168 led to a stepwise decrease in recombinant AOC3 activity. The strongest inhibition was 169 observed at a molar AOC3/ZAG ratio of ~ 1:1 (25 ng ZAG) (Fig. 2, C). GST alone did not show 170 any inhibitory effect. Next, the mechanism of inhibition was investigated by generating a 171 8 Michaelis-Menten plot, which revealed that Vmax (maximum velocity) decreases, whereas Km 172 (i.e. the Michaelis-Menten constant: substrate concentration at half-maximum velocity) 173 remains almost constant, with rising ZAG concentrations (Fig. 2, D). A Lineweaver-Burk 174 diagram clearly illustrates the difference in Vmax and Km behavior. A constant Km value is 175 represented by the intersection of the function with the y-axis (Fig. 2, E). This suggests that 176 ZAG functions as a highly effective allosteric inhibitor of AOC3. It means that ZAG binds 177 AOC3, but not at the catalytic site, thereby reducing the activity of the enzyme in a non-178 competitive manner. 179 Subsequently, it was tested whether recombinant mammalian ZAG inhibits endogenous AOC3 180 activity as effectively as that of recombinant AOC3. Since AOC3 is expressed on the surface 181 of adipocytes and endothelial cells, differentiated 3T3-L1 adipocytes and human coronary 182 artery endothelial cells (HCAEC) were chosen. As a positive control for inhibition, AOC3 183 activity was blocked by inhibitor LJP1586. To eliminate any non-specific background signals 184 in the cell experiments, assays were performed radioactively. The activity of 3T3-L1-derived 185 AOC3 was effectively reduced as the amount of recombinant ZAG was increased. The addition 186 of 50 µg/ml recombinant ZAG inhibited [ 14 C]-benzaldehyde formation to a similar extent as 187 LJP1586 (Fig. 3, A). This is remarkable since the highest concentration of ZAG used (50 188 µg/ml=~1.2 µM) is nearly tenfold less in molar terms than for the small molecule inhibitor 189 LJP1586 (10 µM). This underpins the highly specific nature of this protein interaction, with 190 similar ZAG concentrations being present in human plasma (~50 µg/ml serum) [50]. 191 Furthermore, the inhibitory effect of recombinant ZAG on HCAEC AOC3 (Fig. 3, B) 192 confirmed the similarity between murine and human AOC3, underlining the crosslinking 193 results obtained with SGBS cell membranes and indicating that the ZAG-AOC3 interaction 194 also plays an important role in humans. Since the inhibitor LJP1586 is designed for murine 195 AOC3, a higher concentration was needed to block human AOC3 of HCAEC origin. Comparing the raw data, 3T3-L1 adipocytes and HCAEC cells showed the same AOC3 197 activity. The tenfold-higher activity of 3T3-L1 adipocytes compared with HCAEC relates to 198 the normalization to mg cellular protein/measurement: 3T3-L1 adipocytes contain much less 199 protein.

200
Since recombinant ZAG inhibits endogenous AOC3, it was asked whether endogenous ZAG 201 could inhibit recombinant AOC3. For this purpose, plasma of C57Bl6 wt mice was collected 202 and rebuffered in 10 mM Tris-HCl, pH 8 (ZAG-pI: ~5.8). Plasma proteins were separated by 203 anion exchange chromatography and eluted by linear NaCl gradient (Fig. 4, A). ZAG-containing 204 fractions were identified by WB and used for activity assays. Fractions C12 and D1 showed a 205 signal between 37 kDa and 50 kDa, which corresponds to murine plasma ZAG (Fig. 4, A). ZAG-206 containing fractions (C12 and D1) were pooled, as were fractions without any ZAG (D3 and 207 D4) as controls, and incubated with 50 ng recombinant AOC3 (Fig. 4, B). The ZAG-positive 208 fractions reduced recombinant AOC3 in a dose-dependent manner. However, the control IEX 209 fractions, which contained no ZAG, enhanced recombinant AOC3 activity in a dose-dependent 210 manner, rather than having the expected neutral effect. This stimulatory effect is probably due 211 to both endogenous AOC3 activity and plasma components. First, murine plasma (except that 212 of AOC3 k.o. mice) contains endogenous amine oxidase activity, which can be blocked by the 213 inhibitor LJP1586 (Supplemental Fig. 1 Fig. 2, C). Second, incubation of recombinant AOC3 with plasma of wt, AOC3 218 k.o. and ZAG k.o. mice enhanced AOC3 activity ~3-fold (Supplemental Fig. 1, D). Therefore, 219 a plasma component found in all three genotypes must be responsible for enhancing AOC3 220 activity. Third, the IEX fractions lacking ZAG (D3 and D4) correspond to the major protein 221 10 peak of the IEX profile, which mostly derives from albumin. Incubating recombinant AOC3 222 with fatty acid-free bovine serum albumin (BSA) also enhanced recombinant AOC3 activity 223 to the same extent as murine plasma (Supplemental Fig. 2, A) lacking ZAG to recombinant AOC3 (Fig. 4, B).

232
To substantiate this finding, the plasma of wt mice and ZAG k.o. mice were compared. ZAG-233 containing fractions (C12 and D1) of wt plasma were identified by WB using α-ZAG antibody   Fig. 5, A). Compared to ZAG, isoproterenol significantly enhanced glycerol release already 246 within the first thirteen minutes. ZAG, GST and LJP1586 showed no such effect. Incubating 247 differentiated 3T3-L1 cells with ZAG (50 µg/ml), GST (50 µg/ml) and LJP1586 (10 µM) for 248 several hours revealed that, although ZAG showed a lipolytic effect, this did not occur until 249 twelve hours ( 250 Fig. 5, B). Therefore, ZAG definitely does not behave as a classical β-adrenergic agonist such 251 as isoproterenol and another mechanism must be involved in ZAG-stimulated lipolysis, most 252 likely involving AOC3. Although it is not well characterized, AOC3 is thought to be involved  Fig. 5, D). Comparing the two assays revealed that histamine, cysteamine, cadaverine and 267 octopamine (trace amine) are converted by AOC3 and stimulate lipolysis to a varying degree. 268 Notably, noradrenaline and octopamine both strongly stimulated lipolysis, but only  C57Bl6 mice was not always the same size (Fig. 7, A1). The notion that this size difference  protein was observed, however, which underlines the notion that size differences depend on N-315 glycosylation events (Fig. 7, B). The N-glycosylation siteknown as the sequonis defined 316 by the amino acid sequence Asn-X-Ser (asparagine-X-serine) or Asn-X-Thr (asparagine-X-317 threonine). X can be any amino acid except proline and the Asn residue serves as the anchor i.e. located on the surface of the cell (Fig. 8, B2). Comparing the wt forms showed that HEK293 361 cell-derived ZAG inhibited AOC3 activity, whereas the Expi293F cell-derived ZAG did not 362 ( Fig. 8, C). Although the less-glycosylated form of Expi293F cell-derived ZAG is very likely 363 to be present with the hyperglycosylated form, its inhibitory potential is strongly reduced.

364
Hence, the inhibitory effect of recombinant ZAG depends on which expression host is used. 365 Furthermore, the various ZAG glycomutants, produced in both HEK293 and Expi293F cells, 366 showed a widely differing ability to inhibit AOC3. Importantly, the loss of all glycosylation 367 sites (Δ3-ZAG) led to the same inhibitory potential in both forms of the protein, whether 368 produced in HEK293 or Expi293F cells, which confirms the impact of aberrant glycosylation.

369
Since the ZAG molecular weight was not consistent in all plasma samples, plasma of different 370 mouse strains was collected. WB analysis of mouse plasma of different mouse strains did not 371 show a homogenous pattern (Fig. 8, D). Plasma ZAG of DBA and FVB mice showed a more 372 disperse pattern as observed when overexpressing ZAG in Expi293F cells. the same inhibitory potential as the highly selective inhibitor LJP1586 (Fig. 2, F and G). Regarding octopamine-stimulated lipolysis in the presence of LJP1586 and ZAG (Fig. 6) [49], which is suggested to serve as a redox switch,  Fig. 3). In the presence of AOC3, only wt ZAG 516 shifts to a higher molecular weight, whereas in the absence of AOC3 it does not. By contrast, 517 incubation with H2O2 induces a shift in wt ZAG, irrespective of whether AOC3 is present. This 518 could hint at an oxidation-dependent oligomerization of ZAG, influencing AOC3/ZAG and/or 519 ZAG/ZAG protein-protein interaction, in which glycosylation plays an additional role. ZAG 520 oligomerization could serve as a self-regulatory mechanism, and explain why a complete 521 inhibition of activity was never observed at an equimolar ratio of both proteins (Fig. 2, C), as 522 well as why ZAG loses its lipolysis-stimulatory effect (Fig. 6).

523
Besides H2O2, NH3 might also serve as a signaling molecule. Compared with H2O2, less is 524 known about its function in this context. NH3 is known to stimulate autophagy, playing an 525 important role in energy metabolism in tumor cells [123]. In summary, H2O2 and perhaps also 526 NH3 may have currently uncharacterized effects on AOC3 activitywith or without ZAG 527 modulationthat interfere with signaling pathways. This represents a challenge to researchers 528 to identify physiological compounds serving as substrates for AOC3.

529
During this study, many different expression hosts were tested to find a way to express both 530 AOC3 and ZAG in sufficient, biologically active quantities. Specifically, a surprising 531 difference between HEK293 and Expi293F was observed. Both derive from the same attached 532 cell line, but the latter has been adapted to grow in suspension. Compared with HEK293 cells, 533 overexpression of ZAG in Expi293F cells results in a hyperglycosylated andto a lesser extent 534 hypoglycosylated form (Fig. 8, A2). Different glycosylated forms of ZAG were previously 535 identified by isoelectric focusing and are found in plasma, amniotic fluid, saliva and tears [64].

536
The carbohydrate content of human plasma-derived ZAG makes up to 12-15% of total mass had different molecular weights in most tissues and plasma [47]. This is in line with the 540 observation that plasma ZAG from different mouse strains also shows no coherent pattern (Fig.   541 8, D). Tunicamycin, BAGN and PNGase F treatment of purified ZAG proteins confirmed that 542 size differences originate from N-glycosylation. Strikingly, the addition of 500 µM CoCl2a 543 hypoxia mimetic that stabilizes the transcription factor HIF1-αreverses this effect (Fig. 8,   544 A3 and A4). Glycosylation of proteins is highly variable among individuals and is influenced 545 by oxygen levels. For instance, hypoxia has been shown to reduce uridine diphosphate N- Expi293F cells and shows markedly reduced inhibition of AOC3 activity (Fig. 8, C),    shaker. Cells were grown in Expi293™ Expression Medium (Gibco) and split 1:10 on reaching 637 a density of 5x10 6 cells/ml.

638
All cells were grown in a CO2-controlled incubator with a relative humidity of 90% at 37°C.