Effects of Lipid on the Formation of Heterocyclic Aromatic Amines and Advanced Glycation End Products in Thermally Processed Fish

ABSTRACT The formation of hazardous substances, heterocyclic aromatic amines (HAAs) and two representative advanced glycation end products (AGEs), Nε-carboxymethyl-lysine (CML) and Nε-carboxyethyl-lysine (CEL) in five kinds of fish with different fat contents during thermal processing was investigated. The results indicated that HAAs, AGEs produced is related to the lipids in varying degrees. The variation of key precursors were examined to explain the relationship between fat oxidation and molecular pathway of HAAs and AGEs formation. The current research may help us better understand the overall generation of hazardous substances in food systems, and provide new ideas for formulating prevention in future production.


Introduction
Fish is an excellent source of high-quality protein, minerals, vitamins, and omega-3 fatty acids (including eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA)), which make a great contribution to the daily dietary health of humans (Michael et al. 2019;Prato and Biandolino 2015).On the other hand, fish may also contain many risks such as microbial contaminants, toxins, or parasites.High-temperature thermal processing is an important link in production.But it is worth noting that protein rich fish after heat processing is also associated with an increased risk of dietary intake of nonnegligible hazardous substances such as HAAs and AGEs.Consequently, how to reduce the generation of hazardous substances in high temperature processing production is an urgent issue.
A potentially mutagenic and carcinogenic compound, heterocyclic aromatic amines (HAAs), is derived from the Maillard reaction of proteins or amino acids during high-temperature cooking (generally over 150 ℃) and has become one of the hot issues in the field of food safety in recent years.To date, over 30 types of HAAs have been found and identified in various thermally processed meat products (Barzegar et al. 2019;Gibis 2016).The clear formation pathway of HAAs is complicated and has not been completely elucidated.At present, it is generally thought the formation of HAAs may be closely related to the precursors including creatine, creatinine, amino acids, proteins, pyrazines, and glucose (Jackson and Hargraves 1995;Jägerstad et al. 1983).Advanced glycation end products (AGEs) are many sorts of complex compounds produced during the last stage of the Maillard reaction (nonenzymatic reactions between the carbonyl groups of reducing sugars and the amine groups of proteins).In recent years, people have realized that AGEs are the cause of many chronic diseases, such as diabetes, obesity and so on (Li et al. 2015;Singh et al. 2001).Two typical AGEs, Nε-carboxymethyl-lysine (CML) and Nε-carboxyethyl-lysine (CEL) are stable AGEs in the food system.Studies have shown that the formation of CML and CEL in the food system mainly has the following findings.First of all, Schiff base, a kind of intermediate product of the Maillard reaction will be rearranged by Amadori to form rearranged products and then auto-oxidatively decomposed to form CML and CEL.Glucose in food will also react with lysine to directly form the Schiff base and cleave to form glyoxal, which will continue to react with lysine to form CML and CEL.In addition, glucose or lipid may also autoxidize to form glyoxal and then react with lysine.Therefore, the major precursors of CML and CEL are lysine, glyoxal, and Schiff base (Lin et al. 2022).Nevertheless, despite the AGEs formation having been studied for many years, its exact formation mechanisms are still controversial, especially in meat thermal processing.
Current research on the mechanism of the generation of the two hazardous substances is limited to probing into the precursors; however, considering precursors only is suboptimal.To further inhibit the generation of hazardous substances during thermal processing production, it is necessary to determine the relationship between the other constituent and their generation.Lipids are an important component in food, and the lipid content, composition, and oxidation degree are likely to be closely related to the generation of hazardous substances.Some researchers have previously examined the effect of fat on HAAs formation in model systems and reported that fat can improve the generation of HAAs and there is an optimal level of promotion (10% in their study) (Knize et al. 1985).Fatty acid composition is also inextricably linked to the generation of HAAs.A few studies have claimed that there were less HAAs detected in fish compared to pork, beef, or mutton under similar conditions, which may be due to the high PUFAs in fish.But it has also been reported that an increased proportion of polyunsaturated fatty acids in barbecued meatballs with the same fat content instead increases the content of HAAs.Secondary products produced by lipid oxidation, such as aldehydes and ketones, can react with proteins, amino acids, and other chemical components in meat products to generate carbonyl derivatives, protein-protein complexes, protein-lipid complexes, and reaction pathways affecting protein oxidation and degradation of nonenzymatic browning reactions.These products are closely related to the formation of heterocyclic amines and AGEs.Some studies have shown that a fat oxidation compound, 4-oxo-2-nonenal, participates in Strecker degradation reaction and promotes the formation of PhIP under the joint action of amino acids (Gibis 2016;Wang et al. 2018).Some studies have reported that lipid oxidation can accelerate the formation of CML and CEL with the seduction of malondialdehyde (MDA), and as other important di-carbonyl compounds, glyoxal has been proven an important precursor of CML and CEL (Zhu et al. 2021).However, current research is still insufficient, and comparisons of hazardous substances among meats from different species are less studied.Anoploma fimbria (Goetz et al. 2021), sardines (Luan et al. 2018), and raja porosa (Pan et al. 2016) are rich in omega-3 fatty acids, which can reduce cholesterol and prevent cardiovascular and cerebrovascular diseases.Spanish mackerel (Wu et al. 2021) and eel (Belpaire et al. 2011) are rich in protein, fat, vitamins, and trace elements.Therefore, five kinds of marine fish that vary widely in fat content and readily available locally (anoploma fimbria, raja porosa, Spanish mackerel, sardine, eel (Anguillidae)) were taken as the research objects.The generation of HAAs and AGEs (CML, CEL) were determined in our study, and the changes of basic composition, fat content, contents of four main precursors of HAAs and AGEs (creatine, creatinine, reducing sugar, glyoxal), fatty acid composition, and lipid oxidation degree of fish with and without roasted was detected from the multidimensional aspects such as TBARS, POV, and acid valence in order to establish the connection between the contents of two kinds of hazardous substances and it.Furthermore, we have focused on determining the relationship between the generation of hazardous substances and fat content, lipid composition, and the degree of lipid oxidation, to gain a deeper insight into providing ideas and references for further elucidation of their possible generation mechanism.

Preparation of roasted fish
Fresh eel, Spanish mackerel, raja porosa, anoploma fimbria, and salmon had the scales, skin, head, tail and other parts of the fish without meat removed; samples were sliced and crushed into surimi after being transported from the market.Fifty g of surimi and 2 g soybean oil were mixed and shaped to be a patty in a petri dish (Φ8.5 cm × 1.2 cm, n = 3 for each experimental treatment).All fish patties for each kind of fish were divided into two groups individually and baked in preheated ovens (Germany, RATIONAL) at 250°C and 220°C for 25 min per side, respectively.The patties were cooled at room temperature naturally after being roasted.All samples were crushed and stored at − 20°C for subsequent analysis.

Determination of HAAs
A liquid chromatography-mass spectrometry (LC-MS) (Agilent, 1290-6460) system with gradient elution was used for the chromatographic separation to determine the content of seven kinds of HAAS in roasted fish species.The pretreatment assay of fish samples was based on our previous studies (Li et al. 2022).Patties were homogenized, and the sample (2 g) was mixed with 1 mL deionized water and shaken for 10 min.Then, 9 mL acetonitrile solution (1% acetic acid contained) was added and continued to shake for 10 min.Four g anhydrous magnesium sulfate and 1 g sewage sodium acetate were added and manually shaken vigorously for 1 min, then centrifuged for 10 min (5167.84g).Six mL supernatant was added into the centrifuge tube, with 1 min oscillation and centrifugation for 5 min at 5167.84 g.Then, 1 mL supernatant was dried with Termovap Sample Concentrator, and 1 mL methanol was added for redissolving, and samples were filtered with a 0.22 μm membrane filter.
The acquisition mode of the instrument was ESI+.An Agilent C18 column (2.1 mm x 100 mm, 1.8 μm) was selected, with a flow phase of 5 mmol/L ammonium acetate aqueous solution containing a gradient of 0.1% formic acid (A)/acetonitrile (B) and a flow rate controlled at 0.3 mL/min.The column temperature was controlled at 30°C, and the separation gradient was set at 95% A and 5% A, 0 min to 1.5 min, 90% A and 10% B, 1.5 min to 10 min, 70% A and 30% B, 10 min to 15 min, 15 min to 20 min to 99% A and 1% B, for 1 min.The initial components of 95% A and 5% B were eventually recovered from 21.1 min to 22 min, respectively.

Determination of CEL and CML
A fish sample of 0.1 g was mixed with appropriate amount of physiological saline, and the homogenate was mashed and then centrifuged at 1291.96 g for 10 min.Nε-Carboxyethyl lysine (CEL) and Carboxymethyl lysine (CML) were analyzed by Fish Nε-Carboxyethyl lysine ELISA Kit and Fish Carboxymethyl lysine ELISA Kit (Enzyme-linked Biotechnology Co., Ltd., Shanghai, China) (Li et al. 2022).All the samples were tested.The final 'stop solution' changed the color from blue to yellow, and the intensity of the color was measured at 450 nm using a spectrophotometer.Standard linear regression curves (0, 75, 150, 300, 600, 1200 ng/mL) were drawn according to the instructions to calculate the concentration.

Determination of proximate composition and physicochemical properties
All fish patties were weighed before and after roasting to calculate the cooking loss.The pH values were determined by a digital pH meter (intelligent sensor, Shanghai) according to the method of GB50209.237-2016,water content was determined in GB 5009.3-2016, the crude fat content was tested by the Soxhlet extraction method, and the Kjeldahl method was used to measure the content of protein.

Determination of lipid oxidation
Lipid oxidation of samples was determined by thiobarbituric acid reactive substances (TBARS) assay.One g of minced sample was added to 5 mL of thiobarbituric acid solution (15% trichloroacetic acid, 0.375% 2-thiobarbituric acid, and 0.25 mol/L hydrochloric acid).The mixture was boiled for 20 min, cooled, and centrifuged at 4°C, 9187.28 g for 20 min.The absorbance of the supernatant was measured at 532 nm and 600 nm.

Determination of acid value
The method of determining acid value was based on GB5009.229-2016.The fish oil to be tested was extracted for 12 h with petroleum ether twice the volume of the samples.Potassium hydroxide standard solution was calibrated with potassium hydrogen phthalate, and the concentration resulting from the calculation was noted as C.Then, 50 mL of deionized water (contained two drops of phenolphthalein indicator) was titrated with potassium hydroxide standard solution against to pink, and the solution consumption volume was noted as V2.The extracted fish oil was dissolved in 50 mL of ether-isopropanol solution and shaken after adding 3-4 drops of thymol blue indicator.The mixture was titrated with potassium hydroxide standard solution to blue, and the endpoint was noted as V1.Acid valence was calculated from the following equation:

Determination of peroxide value (POV)
Extracted fish oil samples (0.1 g) were weighed accurately and mixed with 3 mL isooctane vortex, and triplicate 50 μL solution was mixed with 1.45 mL of prepared methanol-N-butanol (2:1), 10 μL ammonium thiocyanate (1 g/mL), and 10 μL fresh ferrous chloride (0.81 g barium chloride and 1 g ferrous sulfate reacted in 0.5 mol/L HCl), reacted in the dark for 20 min and measured the absorbance value at 510 nm (Wang et al. 2022).

Determination of precursors in roasted fish patties
One g of fish samples were pretreated with 20 mL of 30 g/L trichloroacetic acid solution and degreased with ether.The supernatant was filtered by filter paper.Then, 8 mL extraction solution was mixed with 4 mL ether and stood for 10 min to complete degreasing.Four mL of the bottom sample was mixed with 2 mL butanedione and 2 mL 25 g/L 1-naphthol sodium hydroxide solution and soaked in a constant temperature water bath at 40°C for 5 min.The absorbance at 520 nm was measured and sequentially bring the absorbance value into the standard curve to calculate the creatine concentration.The creatinine content was determined by a peroxidase method kit (Nanjing Jiancheng Bioengineering Institute, China), and the detection method of creatinine content was adopted using the protocols reported by Dong et al. (2020).
All the samples were pretreated with zinc acetate solution and potassium ferrocyanide solution.Water extraction and centrifugation were repeated, and the glucose solutions to be tested were obtained after diluting with distilled water to 100 mL.Glucose content was determined by the DNS method, and the absorbance of solutions was measured at 540 nm.The results were analyzed according to the standard curve (0, 0.2, 0.4, 0.6, 0.8, 1.0 mg/mL).
The method of glyoxal determination was with modifications based on S.H. Huang et al. (2022).Then, 1.0 g sample was minced, homogenized with 9 mL pH 7.2 phosphate buffer at 10 000 × g for 1 min (2 × 30 s with a 10 s interval, 4°C), and centrifuged (10 000 × g, 10 min, 4°C).A sample of 0.5 mL supernatant was mixed with 1 mL sodium acetate solution (15 g/L) and 2 mL hydroxylamine hydrochloride (2 g/L) and reheated by the 50°C constant temperature water bath for 20 minutes.After cooling, the volume was fixed with distilled water to 50 mL.The absorbance of the supernatant was measured at 233 nm.

Determination of fatty acid in roasted fish patties
Five fish samples were hydrolyzed with hydrochloric acid.Diethyl ether petroleum ether mixture was added and evaporated to adequately to obtain the fat extract.Two mL of 2% sodium hydroxidemethanol solution and the fat extract were mixed and heated in a water bath at 85°C for 30 min, and then 3 mL of 14% boron trifluoride methanol solution was added to continue the water bath for 30 min.After cooling to room temperature, it was mixed with 1 mL of n-Hexane and shaken for 2 min.After allowing it to stand for 1 h, 100 μL supernatant was taken to dilute with n-hexane to 1 mL and filtered by a 0.45 μm organic membrane filters (Milex, USA).
Fatty acid methyl esters were separated in a Trace 1310 ISQ GC system (Thermo, Shanghai, China) fitted with a TG-5 MS chromatographic column (30 m × 0.25 mm × 0.25 μm) selected in the experiment.The instrument testing procedure was set as follows: the gas-phase injection port temperature was set to 290°C; the carrier gas flow rate was 1.2 mL/min; the splitless mode was selected; and the temperature program was firstly 80°C held for 1 min, raised to 200°C at a rate of 10°C/min, continued to raise to 250°C at a rate of 5°C/min, and finally raised to 270°C at a rate of 2°C/min and held for 3 min.The temperature between the ion source and the transfer line of the mass spectrum was set to 280°C, the solvent delay was 5.00 min, the scan range was 30 ~ 400 amu, and the ion source selected the EI source with an energy of 70 eV.The type and content of all fatty acid methyl esters were determined by checking the retention time and calculation of peak areas of the sample fatty acid methyl esters with the retention time of 35 fatty acid methyl esters standards, respectively, and the content of fatty acids was calculated according to the following formula: C(mg/L): concentration of fatty acid methyl esters in the test solutions; V(mL): volume of the test solutions; k: conversion coefficients of fatty acid methyl esters to fatty acids; m(g): mass of the samples.

Statistical analysis
All samples were analyzed in triplicate using mean ± standard (SD).Statistical analysis used unidirectional variance analysis (ANOVA) to evaluate the significance of differences between treatment groups (*p < 0.05) (IBM SPSS Statistics 26).

Quantification of HAAs, CML, and CEL in roasted fish at different temperature
The relationship between temperature and HAAs content has been confirmed to be quite close (Juhee and Grün 2006).The standard curves and chromatograms of seven types of HAAs standards are shown in Figure S1.The hazardous substances content of five kinds of fish samples treated with 220°C and 250°C are given as ng/g in Table 1, and Figures 1 and S2 shows their chromatograms.With the temperature rising, the amount of HAAs increased significantly.At 220°C, few HAAs were detected and Norharman predominated.In the meantime, IQx, MeIQ, and PhIP were not detected in all fish samples.It is noteworthy that all seven HAAs were detected in samples at 250°C.This indicates that the formation of IQx, MeIQ, and PhIP requires higher temperature conditions.The HAAs content of all five fishes increased significantly with increasing temperature, which is consistent with previous reports (Li et al. 2022;Lin et al. 2016).Among them, the concentration of HAAs in roasted raja porosa changed the most (from 5.8998 ± 0.0078 to 34.4763 ± 0.0534), and the least was anoploma fimbria (from 2.9811 ± 0.0074 to 6.5597 ± 0.0158).Costa et al. (2009) studied the production of HAAs in sardines and salmon after high temperature thermal processing and summarized the results of other studies (sardine, eel, cod, herring, and trout), showing that the production species and contents of HAAs in different fish species are significantly different.It was demonstrated in our study that 4,8-DiMeIQx was consistently not detected in anoploma fimbria and Raja porosa, and 8-MeIQx was also consistently not detected in sardine and Raja porosa.The phenomenon fully illustrates the dramatic effect on the gap in muscle composition on the species and number of HAAs production.
The data shows that the content of CML and CEL in raw meat is much higher than the value after roasting, and the content after roasting is not changed significantly.The highest content of CML and CEL is the raw Spanish mackerel group (42.48 ± 1.07 ng/g), and the raw anoploma fimbria group (37.47 ± 1.91ng/g), respectively.This may be due to the complete denaturation of protein and the complete breaking of disulfide bond after high temperature thermal processing, which triggers CML and CEL to stay at low level stability.In addition, refrigerated storage will also lead to an increase in CML and CEL content.

Physicochemical characteristics analyses
Table S2 shows the cooking loss, moisture content, lipid content, and pH of different fish samples.In this study, five kinds of fish with different fat contents were selected, and their fat content was tested (eel: 11.02 ± 0.565%, sardine: 0.26 ± 0.01%, Spanish mackerel: 2.18 ± 0.05%, raja porosa: 0.35 ± 0.04%, anoploma fimbria: 20.00 ± 2.646%).Under high-temperature thermal processing conditions, the different contents of water, myofibril, and lipids in fish meat will lead to the loss of various components to varying degrees during the cooking process.The amount of loss and changes in water content represent the water retention of meat, which is closely related to the production of HAAs.As can be seen from the results, raja porosa has the worst water retention property, the water content of which was the highest, and directly drops to the lowest after roasting (from 77.87 ± 0.22% to 11.03 ± 0.58%).The anoploma fimbria has the strongest water retention, and the water content did not change significantly after roasting (from 63.50 ± 2.66% to 53.03 ± 2.32%), which is consistent with the rule of HAAs generation.In the heat map of the contrast between raw meat (Figure 2a), we can see that fat has a strong effect on water conservation.Some scholars have studied the effect of pH on the formation of CML in the methylglyoxal-glucose-amino acid model system.They observed that lower pH inhibited the formation of CML.In our study, raja porosa has the lowest pH value (6.9400 ± 0.0116) and the lowest CML and CEL detected.The pH value of Spanish mackerel and anoploma fimbria was the highest (7.5367 ± 0.0088 and 7.4833 ± 0.0067), and the CML and CEL detected were also the highest, which is consistent with the trend.
In the heat map in Figure 2, associations are shown with greater clarity that CML and pH showed a strong positive correlation at various temperatures (r = 0.62; p < 0.05), and HAAs content and moisture are also negatively correlated (r = −0.93;p < 0.001).In the raw group (Figure 2a), fat content is positively correlated with CML and CEL (r = 0.61; p < 0.05), while after roasting, fat content is negatively correlated with both CML and CEL (Figures 2b,c; r = −0.75;p < 0.01).This difficult to explain phenomenon may be because the fat content of fish patties during thermal processing conditions is enriched, and at the same time, also hindering the generation of CML and CEL to some extent.There have been studies claiming no apparent relationship between protein or lipid concentrations and CML/CEL levels in raw or heated pork offal (Niu et al. 2022).

Change of precursors in roasted fish patties
Many studies have confirmed that creatine, creatinine, and glucose are precursors of HAAs, glucose, and glyoxal are two kinds of the main precursors of AGEs.The concentrations of precursors (creatine, creatinine, glyoxal and glucose) in roasted fish are shown in Table S3.
The overall trend of creatine shows that there is a certain accumulation at 220°C, while when heated to 250°C, a large amount of creatine is consumed to generate HAAs.Taking raja porosa with the largest amount of HAAs as an example, we can grasp that it is different from other fish.With the increase of roasting temperature, the poor water retention of raja porosa may result in a large amount of creatine loss (Wyss and Kaddurah-Daouk 2000) (from 1460.4768 ± 12.6304 to 900.5569 ± 12.7287) in the system converted into creatinine (from 1.5867 ± 0.0741 to 2.5991 ± 0.1532), which provides favorable conditions for a large number of HAAs.It can also be reflected from the creatine creatinine data of Spanish mackerel and anoploma fimbria with low HAAs content that their creatine creatinine consumption is significantly lower than that of other fish.
Glucose and amino acids are the main precursors of CML, and myosin is the key point.Over time, myosin is oxidized and becomes a partially unfolded conformation.From the data, we can see that the changing trend of glucose content in the sardine with the highest content of CML and CEL is the most obvious, and the raja porosa with the lowest content of CML and CEL has the glucose content with minimum fluctuation.Figure 2a also shows that glucose content is strongly negatively correlated with AGEs (CML: r = −0.89;p < 0.001; r = −0.53;p < 0.05), indicating that glucose, as a precursor, participates in a reaction that is consumed stepwise to generate CML and CEL.After being roasted at 220°C, the correlation between glucose and CML in fish patties disappeared.Furthermore, none had a correlation after being roasted at 250°C.The result may be due to the content of glucose in fish patties increasing dramatically during the thermal process, while CML and CEL could not be generated in large amounts under the influence of other factors, thus showing as gradually losing the correlation at the numerical level.A recent study showed that the formation of CML and CEL may not be related to the degree of oxidation but is closely related to the induction of aldehydes such as glyoxal and malondialdehyde (Fang et al. 2022).It was suggested that the lotion will demulsify and separate oil at 30-50 mmol/L glyoxal concentration.In this process, the interaction between glyoxal and lotion protein is the main reason to promote the formation of aging, rather than oxidation.Glyoxal may promote the formation of CML and CEL through direct interaction with muscle protein lotion (similar to the raw fish system in our study); thus, the CML and CEL content of all raw groups is high.This research shows that the oxidation of lipids and proteins under high aldehyde content is not closely related to the formation of AGEs.When the concentration of glyoxal exceeds a certain concentration, the oxidation content is stable, while the content of CML and CEL gradually decreases.This is consistent with our test results; for example, the TBARS value (MDA content, 1.1959 ± 0.0398) of the raja porosa with the lowest content of CML, CEL, and the glyoxal content (0.9023 ± 0.0032) are both the highest.

Correlation between lipid changes and hazardous substances formation
TBARS, acid value (AV), and POV can reflect the oxidation degree of samples from different perspectives.As shown in Table 2, not all fish species had trends in TBARS that were temperature dependent.This may be due to the risk of degradation of MDA (symbol of TBARS), under specific conditions at high temperatures, which shows an inexplicable trend in different fish meat systems.
POV represents the number of peroxides formed by the oxidation and opening of double bonds of unsaturated fatty acids in lipids, and AV represents the degree of direct oxidation of lipids to low-grade aldehydes and ketoacids after enzymatic hydrolysis.It can be seen from the overall trend of POV and AV in the table that the degree of oxidation of fish oil increases significantly with the temperature increase.According to the synthetical analysis of the heat map (Figure 3) and data (Table 2), among the five kinds of fish, sardines (POV from 16.8643 ± 0.2743 to 33.7732 ± 0.1819; AV from 0.6864 ± 0.0274 to 1.0574 ± 0.0182) have the highest degree of oxidation after roasting, while eels (POV from 14.6602 ± 0.0030 to 21.4159 ± 0.3575; AV from 0.2142 ± 0.0036 to 0.4297 ± 0.0074) have the lowest degree of oxidation.The difference between AV and POV of Spanish mackerel and anoploma fimbria is obvious, which may be due to the high content of unsaturated fatty acids in fish oil and the diverse and complex process of oil oxidation.After the double bond is oxidized to form peroxide, it is further oxidized, which has a certain impact on the results.As can be seen from Figure 3a, for fish patties roasted at 220°C, the positive correlations between TBARS and the content of hazardous substances were all significant (HAAs: r = 0.72; p < 0.01; CML: r = 0.87; p < 0.001; CEL: r = 0.73; p < 0.01), POV is  positively correlated with CEL (r = 0.67; p < 0.01), and AV is negatively correlated with HAAs (r = −0.86;p < 0.001).However, the relationship disappeared during treatment at 250°C (Figure 3b), but POV was still positively correlated with CEL (r = 0.72; p < 0.01).This condition may due to hydroxyl radicals (•OH) and lipid peroxyl radicals (LOO-) by lipid oxidation, which can promote the generation of CML and CEL in the Maillard reaction model system (Lu et al. 2023;Zhu et al. 2021).Other scholars have controversial interpretations about the generative relationship of TBARS with CML and CEL, so the mechanisms formed behind this need further inquiry (Liu et al. 2022;Soladoye et al. 2017;Weber et al. 2008).
We also made a brief exploration of the basic fatty acids.GC-MS chromatograms of 35 fatty acids are shown in Figure S3.It can be seen in Table 3 that anoploma fimbria patties have the highest content of unsaturated fatty acids (UFA), and raja porosa patties and sardine patties have the lowest content of UFA.Coincidentally, raja porosa patties have the highest content of HAAs, and anoploma fimbria patties have the lowest content of HAAs.It was found that barbecued meatballs formulated with high proportion of sheep tail fat enriched with linoleic and polyunsaturated fatty acids had higher levels of HAAs after cooking significantly, which might be due to the unsaturated fatty acids promoting the generation of HAAs (Oz 2021).These results were in direct contrast to ours.This may be due to the fact that the gap in the composition of meat among different species has a more significant effect on the variety and quantity of HAAs produced.

Conclusion
In conclusion, this study explored the differences in the formation of hazardous substances in fish with different fat content after baking under different high-temperature thermal processing conditions and conducted research and analysis respectively from the aspects of fat content, lipid composition, and the degree of lipid oxidation.The results showed that the types and contents of HAAs produced by fish with different fat content after baking under the same conditions were significantly different.The HAAs content of all five fishes increased significantly with increasing temperature, and the production species and contents of HAAs in different fish species are significantly different.The concentration of HAAs in roasted raja porosa changed the most (from 5.8998 ± 0.0078 to 34.4763 ± 0.0534), and the least was in anoploma fimbria (from 2.9811 ± 0.0074 to 6.5597 ± 0.0158).The total amount of CML and CEL is also related to the characteristics of lipids in varying degrees.After roasting, fat content is negatively correlated with both CML and CEL.We also studied the changes of precursors of hazardous substances in the thermal process and tried to explain the relationship between fat properties and the molecular pathways of HAAs and AGEs formation in muscles of different species comprehensively.The lipid oxidation degree of samples, including TBARS, AV, and POV is correlated with HAAs and AGEs in different fish meat systems.In general, the data in this study would provide a theoretical basis and valuable guidance for understanding the effects of lipid oxidation on the generation of HAAs and AGEs in various fish food systems and provide new ideas for the development of prevention and control measures for hazardous substances in future production.

Disclosure statement
No potential conflict of interest was reported by the author(s).
detected (below the limit of detection); Results were expressed as mean ± standard deviations.a-d = Different letters indicated significant differences (*p < 0.05) between samples for different treatments in the same column.

Figure 2 .
Figure 2. Heat map of the relationship between hazardous substances and physicochemical properties and precursors in five kinds of fishes with no treatment (a), roasted at 220°C (b), roasted at 250°C (c).

Figure 3 .
Figure 3. Heat map of the relationship between hazardous substances and lipid oxidation degree changes in five kinds of fishes before and after roasting at 220°C (a), roasted at 250°C (b).

Table 1 .
Content (ng/g) of hazardous substances in five kinds of roasted fish.

Table 2 .
Lipid oxidation degree in five kinds of fish treated with different conditions.
bResults were expressed as mean ± standard deviations.a-d = Different letters indicated significant differences (*p < 0.05) between samples for different treatments in the same column.

Table 3 .
The content of fatty acid composition in five kinds of fish patties (mg/g).detected (below the limit of detection).mg/g represents the converted fatty acid composition in per gram of fish.