Dietary magnesium deficiency impaired intestinal structural integrity in grass carp (Ctenopharyngodon idella)

Grass carp (223.85–757.33 g) were fed diets supplemented with magnesium (73.54–1054.53 mg/kg) for 60 days to explore the impacts of magnesium deficiency on the growth and intestinal structural integrity of the fish. The results demonstrated that magnesium deficiency suppressed the growth and damaged the intestinal structural integrity of the fish. We first demonstrated that magnesium is partly involved in (1) attenuating antioxidant ability by suppressing Nrf2 signalling to decrease antioxidant enzyme mRNA levels and activities (except CuZnSOD mRNA levels and activities); (2) aggravating apoptosis by activating JNK (not p38MAPK) signalling to upregulate proapoptotic protein (Apaf-1, Bax and FasL) and caspase-2, -3, -7, -8 and -9 gene expression but downregulate antiapoptotic protein (Bcl-2, IAP and Mcl-1b) gene expression; (3) weakening the function of tight junctional complexes (TJs) by promoting myosin light chain kinase (MLCK) signalling to downregulate TJ gene expression [except claudin-7, ZO-2b and claudin-15 gene expression]. Additionally, based on percent weight gain (PWG), against reactive oxygen species (ROS), against caspase-9 and claudin-3c in grass carp, the optimal dietary magnesium levels were calculated to be 770.38, 839.86, 856.79 and 811.49 mg/kg, respectively.

Magnesium is an essential element well known for its role in activating enzymes for nutrition metabolism, energy metabolism and nucleic acid biochemistry in mammals 1 . Emerging evidence has revealed that magnesium deficiency could induce inflammation in human 2 and rat intestines 3 . A recent study demonstrated that inflammation could impair animal intestinal structural integrity 4 . These results indicate that magnesium deficiency might impair animal intestinal structural integrity. Unfortunately, so far, only one study has observed that magnesium deficiency impaired mouse intestinal structural integrity by down-regulating occludin and ZO-1 gene expression 5 . However, this research still lacks a systematic approach to animal intestinal structural integrity, and it did not investigate the underlying mechanisms. Therefore, it is imperative to explore the effects of magnesium deficiency on intestinal structural integrity and to conduct deeper examination on the molecular mechanisms in animals.
In fish, intestinal structural integrity is influenced by cellular structural integrity, which can be impaired by cell apoptosis and oxidative damage 6 . Chen et al. 7 found that antioxidants could attenuate oxidative damage in grass carp intestine. Moreover, another study observed that cell apoptosis depended on apoptosis -related proteins of the caspase family (caspase-2, -3, -7, -8 and -9) in mammals 8 . In fish, antioxidants and apoptosis-related proteins are deeply dependent on regulation by Nrf2 9 and JNK 10 , respectively. So far, the fragmentary research of oxidative damage (only detecting MDA and ROS) and cell apoptosis (only detecting caspase-3) in animals has been focused on the liver, kidney, heart, brain, muscle, thymus and spleen [11][12][13][14] . However, animal intestines differ from these tissues and organs in terms of oxidative damage and cell apoptosis, and here, we list some of the differences. First, during normal aging of both humans and animals, some postmitotic tissues can be renewed by cell apoptosis in Oxidative stress parameters in the intestines of grass carp. In Table 2, we can clearly find the effects on the antioxidant related substances in grass carp intestines with magnesium treatment. The content of MDA in grass carp intestines decreased as the magnesium level rose to 861.67 mg/kg and increased significantly afterward (P < 0.05). Meanwhile, the highest PC contents were observed in intestines of grass carp fed a dietary magnesium level of 73.54 mg/kg. Additionally, the ROS content decreased to the lowest level in grass carp MI and PI as the magnesium levels rose to 861.67 mg/kg and in the DI of this fish as the magnesium level rose to 691.55 mg/kg, increasing dramatically thereafter (P < 0.05). On the contrary, the GPx and GST activities increased significantly (P < 0.05) in grass carp intestines when the magnesium levels rose to 861.67 mg/kg and thereafter decreased significantly (P < 0.05). The highest ASA activities were found in the PI of grass carp fed a dietary magnesium level of 861.67 mg/kg and in the DI and MI of this fish fed a dietary magnesium level of 691.55 mg/kg. The AHR activities and GSH contents increased significantly in grass carp MI and PI as the magnesium level rose to 691.55 mg/kg and in the DI of this fish as the magnesium level rose to 861.67 mg/kg (P < 0.05), thereafter decreasing. Additionally, fish fed a dietary magnesium level of 861.67 mg/kg showed maximum activities of T-SOD and MnSOD in grass carp intestines. Fish fed dietary magnesium levels of 861.67 mg/kg and 691.55 mg/kg showed the maximum CAT activities in the MI and in the DI and PI, respectively. The GR activities increased significantly (P < 0.05) in grass carp MI and PI at a magnesium level of 861.67 mg/kg, and in the DI of this fish at a magnesium level of 691.55 mg/kg, thereafter decreasing significantly (P < 0.05). Interestingly, magnesium did not influence the CuZnSOD activities in grass carp intestines.
Observation of the effects on antioxidant-related gene expression also need mentioning (Fig. 2). The gene expression of GSTP2, GPx1a and GSTO1 in grass carp intestines reached to a peak as the magnesium level rose to 861.67 mg/kg and then decreased. Furthermore, the gene expression of GPx1b, GPx4b and GR was upregulated under magnesium treatment with magnesium levels rose to 691.55 mg/kg in the PI and 861.67 mg/kg in the DI and MI and thereafter plateaued (P > 0.05). The MnSOD, Nrf2 and GSTR gene expression was upregulated in grass carp DI and PI when fish received 861.67 mg/kg, and in the MI of this fish at a magnesium level of 691.55 mg/kg, decreasing afterwards. Simultaneously, the gene expression of other cytokines, namely, CAT, GSTP1, GPx4a and GSTO2 was significantly higher in grass carp intestines in the optimal-magnesium group compared with the magnesium-deficient group (P < 0.05). In addition, the Keap1a gene expression in grass carp intestines decreased as the magnesium level rose to 861.67 mg/kg and plateaued thereafter (P > 0.05). Surprisingly, we found that dietary magnesium did not alter the Keap1b and CuZnSOD mRNA levels in grass carp intestines.
Protein levels of Nrf2 in the intestines of grass carp. The impacts of magnesium on cytosolic Nrf2 and nuclear Nrf2 protein levels in grass carp intestines are shown in Fig. 3. When the magnesium level rose to 861.67 mg/kg, the protein levels of nuclear Nrf2 increased sharply (P < 0.05) in grass carp intestines and then declined significantly (P < 0.05). When the magnesium level rose to 861.67 mg/kg, increased protein levels of cytosolic Nrf2 were found in grass carp intestines, which then decreased significantly (P < 0.05). The mRNA levels of genes related to TJs in the intestines of grass carp. Magnesium had multiple influences on TJs-related genes in grass carp intestines. In Fig. 6, it can be seen that the claudin-b, -12, -11,c and ZO-1 gene expression in grass carp intestines increased as the magnesium level rose to 861.67 mg/kg and decreased thereafter. Among all six groups, fish fed a magnesium-deficient diet showed the lowest levels of claudin-3c gene expression in grass carp intestines. Additionally, the gene expression of occludin and claudin-f was highest in grass carp DI and PI with a magnesium level of 861.67 mg/kg and in the MI with a magnesium level of 691.55 mg/kg and then decreased. However, fish in the magnesium-deficient group showed the highest levels of claudin-15b and claudin-15a gene expression in grass carp intestines. Meanwhile, MLCK gene expressions  Values are means and standard deviations of six replicates. a,b,c,d,e,f Mean values within a row with unlike superscript letters are significantly different (P < 0.05). MDA, malondialdehyde (nmol/g tissues); PC, protein carbonyl (nmol/mg protein); ROS, reactive oxygen species (% DCF florescence); ASA, anti-superoxide anion (U/g protein); AHR, anti-hydroxyl radical (U/mg protein); T-SOD, total superoxide dismutase (U/mg protein); CuZnSOD (U/mg protein); MnSOD (U/mg protein); CAT (U/mg protein); GPx (U/mg protein); GST (U/mg protein); GR (U/g protein); GSH (mg/g protein).

Discussion
In this study, we observed that magnesium deficiency suppressed grass carp (223.85-757.33 g) growth with poor FI, PWG, SGR and FE. It has been accepted that fish growth is related to nutrient metabolism in the fish body 41 . The magnesium nutritional level in the fish body can be reflected by the magnesium concentration in serum and tissues 41 . Our results showed that magnesium deficiency decreased the magnesium concentrations in grass carp serum and intestines.
Additionally, fish growth depends on the intestinal growth 45 . To our knowledge, fish intestinal growth can be reflected by the IL, ILI, IW and ISI 6 . According to our present data, magnesium deficiency depressed the IL, ILI, IW and ISI, suggesting that magnesium deficiency could depress fish intestinal growth. It has been generally accepted that fish intestinal growth depends on the intestinal structure 46 . One study demonstrated that goblet cell hyperplasia could thicken the mucus layer in the human intestine 47 . An excessively thickened mucus layer would block the intestinal absorption function in mammals 48,49 . Our histological results showed that magnesium deficiency caused goblet cell hyperplasia in grass carp intestines, which may partly contribute to the decreased activities of intestinal brush border enzymes (such as Na + , K + -ATPase and AKP). It has been demonstrated that AKP and Na + , K + -ATPase are involved in the absorption of nutrients (such as glucose and amino acids) in animal intestine 50,51 . In animal intestines, goblet cells are associated with the absorption of nutrients (such as glucose) 52 .
In the present study, magnesium deficiency suppressed the Na + , K + -ATPase and AKP activities in fish intestines. We hypothesize that magnesium deficiency might decrease the activities of intestinal brush border enzymes (such as Na + , K + -ATPase and AKP), resulting in goblet cell hyperplasia to maintain the intestinal function of absorbing nutrients, an idea that needs more investigation. Magnesium deficiency-induced suppression of the Na + , K + -ATPase and AKP activities might be related to the physiological functions of magnesium. As is known, magnesium is involved in the active site of AKP in Escherichia coli 53 and of Na + , K + -ATPase in animal kidney cells 54 . These results indicate that the depressed fish growth under a magnesium-deficient diet may be attributed to the suppression of intestinal brush border enzymes and the negative intestinal growth. Undeniably, fish growth is related to the intestinal structural integrity, which depends on cellular and intercellular structural integrity 6 . Hence, it is imperative to study the relationship between magnesium deficiency and the cellular and intercellular structural integrity in fish intestines.  It was previously reported that ROS could induce oxidative damage and that ROS can be eliminated by the antioxidant system in fish 6 . Based on the current results, magnesium deficiency increased the contents MDA, PC and ROS while decreasing the antioxidant enzymes (except CuZnSOD) activities and the non-enzymatic antioxidant (GSH) content in grass carp intestines, indicating that magnesium deficiency increases oxidative damage because of decreasing antioxidant ability in fish intestines. To some extent, the gene expression of antioxidant-related enzymes can reflect the antioxidant enzyme activities in animals 55 . As the results show, magnesium deficiency decreased the antioxidant enzyme mRNA levels (except CuZnSOD) in grass carp intestines. Moreover, our study revealed that the antioxidant enzyme activities (GPx, GST, GR, CAT and MnSOD) had a positive correlation to the enzyme gene expression (Table 3). This evidence suggested that magnesium deficiency may downregulate antioxidant enzymes (except CuZnSOD) gene expression to decrease their activities in fish intestines.
Interestingly, we found that dietary magnesium only enhanced MnSOD (not CuZnSOD) activity and gene expression in grass carp intestines, which may partly be attributed to apolipoprotein A-I (ApoA-I). It was reported that magnesium increased the concentration of ApoA-I in rats livers 56 which only upregulated gene expression of MnSOD (not CuZnSOD) and incaresed MnSOD (not CuZnSOD) protein levels in mouse ID8 cells 57 , supporting our hypothesis.
Antioxidant enzyme gene expression is under Nrf2 signalling pathway regulation in mammals 58 . Nrf2 nuclear translocation could activate the Nrf2 signalling pathway, which could be evaluated by the nuclear Nrf2 protein level in mice 59 . Our results suggested that magnesium deficiency could decrease nuclear Nrf2 protein levels to suppress the nuclear translocation of Nrf2, which may inhibit the Nrf2 signalling pathway in fish intestines. Furthermore, our study showed that these antioxidant enzymes (except CuZnSOD) gene expressions had a positive connection to nuclear Nrf2 protein levels in grass carp intestines ( Table 3), suggesting that magnesium deficiency may downregulate antioxidant enzyme gene expression by inhibiting the Nrf2 signalling pathway. The reasons for magnesium deficiency inhibiting Nrf2 nuclear translocation in fish intestines are as follows. First, the inhibition may be attributed to the de novo synthesis inhibition of Nrf2 by magnesium deficiency. It was reported that de novo synthesis inhibition of Nrf2 may block the Nrf2 nuclear translocation process in humans 60 . A previous study in mice reported that Nrf2 nuclear translocation was closely correlated with its transcriptional and translation levels 59 . Our study found that magnesium deficiency suppressed the total Nrf2 translational level (nuclear and cytosolic Nrf2 protein levels) and suppressed the Nrf2 gene transcriptional level (Nrf2 gene expression) in grass carp intestines, indicating that magnesium deficiency inhibited Nrf2 de novo synthesis to suppress Nrf2 nuclear translocation in fish intestines. The downregulation of Nrf2 gene expression by magnesium deficiency in fish intestines may be attributed to vitamin B 6 content. In rat plasma, magnesium deficiency could decrease the vitamin B 6 content 61 . Our laboratory previous study observed that vitamin B 6 deficiency decreased the Nrf2 gene expression in grass carp intestines 62 . Therefore, magnesium deficiency might decrease the vitamin B 6 content to downregulate the Nrf2 gene expression in fish intestines. Second, the inhibition might be attributed to the upregulation of keap1 gene expression by magnesium deficiency. In mice, Keap1 is a Nrf2-binding protein that prevents the Nrf2 nucleus translocation process by facilitating Nrf2 degradation 63 . The current study found that magnesium deficiency upregulated the Keap1a (not Keap1b) gene expression in grass carp intestines. Our study observed that Keap1a (not Keap1b) gene expression had a negative connection to nuclear Nrf2 protein levels in grass carp intestines (Table 3), indicating that the magnesium deficiency-induced suppression of Nrf2 nuclear translocation may occur partially via up-regulation of Keap1a (not Keap1b) gene expression in fish intestines. Our results above suggested that magnesium deficiency weakened the antioxidant capacity in fish intestines, which was partly associated with the Nrf2/Keap1a (not Keap1b) signalling pathway.
In contrast, we found that dietary magnesium downregulated only Keap1a (not Keap1b) gene expression in grass carp intestines, which may be associated with phospholipids. Gimenez et al. 64 reported that magnesium could increase the phospholipids content in rat blood. Previously, our laboratory found that phospholipids decreased only keap1a (not keap1b) gene expression in juvenile grass carp intestines 7 , supporting our hypothesis.  Over all, we observed that magnesium deficiency decreased the antioxidant capacity and caused oxidative damage in fish intestines, which may be modulated by the Nrf2/Keap1a (not Keap1b) signalling pathway. A previous study found that oxidative damage aggravates cell apoptosis in humans 65 . Hence, we next examined the relationship between magnesium deficiency and cell apoptosis in fish intestines.
In addition, JNK and p38MAPK take part in manipulating cell apoptosis in humans 69,70 . By coincidence, magnesium deficiency upregulated JNK gene expression but did not alter p38MAPK mRNA levels in grass carp intestines. The upregulation of JNK gene expression by magnesium deficiency in fish intestines may be attributed to a potassium deficiency. According to one study in animals, magnesium deficiency could cause potassium deficiency 71 . Potassium deficiency could also elevate the JNK protein level in calves 72 . Therefore, magnesium deficiency might cause a potassium deficiency, upregulating JNK gene expression in fish intestines. Afterwards, our study found that proapoptotic protein (Bax, FasL and Apaf-1) gene expression had a positive connection to JNK gene expression, but antiapoptotic protein (Mcl-1b, Bcl-2 and IAP) gene expression had a negative connection to JNK gene expression in grass carp intestines (Table 3). In summary, all evidence above indicates that magnesium deficiency may aggravate apoptosis in fish intestines, partly depending on the [JNK (not p38MAPK)/FasL/ caspase-8/(caspase-3 and -7)] and [JNK (not p38MAPK)/(Bax, Apaf-1, Bcl-2, Mcl-1b and IAP)/(caspase-2 and -9)]/(caspase-3 and -7)] signalling pathways.
Surprisingly, our study observed that magnesium did not alter p38MAPK gene expression in grass carp intestines, which may be attributed to vitamin D. According to a study of human blood, magnesium could increase the vitamin D content of blood 73 . Our previous research found that vitamin D did not alter p38MAPK gene expression in the enterocytes of fish 74 , supporting our hypothesis.
Moreover, TJs are always on the top of the list for maintaining intercellular structural integrity in human Caco-2 cells 75 , which is important for animal intestinal structural integrity 76 . Thus, an investigation of the relationship between magnesium deficiency and TJs in grass carp intestines as well as underlying signalling pathways is necessary.  TJs (such as occludin, claudins and ZO-1) could regulate the intercellular structural integrity in the sea bream (Sparus aurata) gut 77 . Research in mouse intestinal epithelia demonstrated that claudin-15 is one of the pore-forming proteins that improve epithelial permeability 78 . Based on our results, magnesium deficiency decreased occludin, ZO-1, and claudin-c, -b, -3c, -f, -11 and -12 gene expression, but upregulated ZO-2b (only in PI) and claudin-15b and -15a gene expression in grass carp intestines. One study reported that MLCK could regulate the tight junction permeability in terrestrial animals 79,80 , and the activation of MLCK could decrease TJ gene expression in the bovine brain 34 . Our data indicated that magnesium deficiency enhanced MLCK gene expression in grass carp intestines. The MLCK gene expression enhancement by magnesium deficiency in fish intestines may be attributed to an elevated concentration of TNF-α. It has been reported that magnesium deficiency elevated the concentration of TNF-α in humans 35 . Elevated TNF-α could also upregulate MLCK gene expression in humans 81 . Therefore, magnesium deficiency might elevate the TNF-α concentration to upregulate MLCK gene expression in fish intestines. Our study demonstrated that occludin, ZO-1, and claudin-c, -b, -3c, -f, -11 and -12 gene expression had a negative connection to MLCK gene expression, while claudin-15a and -15b and ZO-2b (only in PI) gene expression had a positive connection to MLCK gene expression in grass carp intestines (Table 3). All evidence above suggests that magnesium deficiency damaged the tight junction function in fish intestines, which occurred partly through MLCK signalling pathway suppression of occludin, ZO-1, and claudin-3c, -11, -b, -f, -c and -12 gene expression, and upregulation of claudin-15a and -15b and ZO-2b (only in PI) gene expression.
Surprisingly, we found that dietary magnesium deficiency increased ZO-2b gene expression only in grass carp PI (rather than DI and MI) and that dietary magnesium had no influence on claudin-7a and -7b gene expression in the intestines of this fish. Several reasonable potential causes for these effects are as follows. First, dietary magnesium deficiency upregulated the ZO-2b gene expression only in grass carp PI (rather than DI and MI), which may be attributed to the zinc in fish intestines. It was reported that magnesium deficiency increased the intestinal absorption of zinc in rats 82 . Our laboratory previous study observed that zinc increased the ZO-2b gene expression only in grass carp PI (rather than DI and MI) 83 , supporting our hypothesis. Second, dietary magnesium did not alter claudin-7a and -7b gene expression in grass carp intestines, which may be attributed to Na + , K + -ATPase in fish intestines. Previously, Alexandre et al. 84 reported that claudin-7 is generally accepted as a channel for Na + in pig LLC-PK1 cells. Moreover, magnesium could activate human blood Na + , K + -ATPase activity 85 , which regulates Na + movement in most higher eukaryotes 86 . Hence, we suggest that dietary magnesium may enhance the Na + , K + -ATPase activity to regulate Na + movement instead of claudin-7, resulting in the observed stable gene expression of claudin-7b and -7a in fish intestines. However, this hypothesis deserves deeper research.
Meaningfully, in this study, there are some innovative discoveries of magnesium beyond the previous knowledge of magnesium. We list these novel discoveries as follows: (1) Previous researches involving the effect of magnesium on oxidative damage in aminals has only focused on the oxidation products (ROS, MDA and PC) and antioxidant enzymes (SOD, GST, GPX and CAT) [87][88][89][90][91] . However, apart from the investigation of oxidation products (ROS, MDA and PC), antioxidant enzymes (SOD, GR, GST, GPX and CAT) and non-enzymatic antioxidants (GSH), we studied the mRNA levels of genes corresponding to antioxidant enzymes (CuZnSOD, MnSOD, CAT, GR, GPx1a, GPx1b, GPx4a, GPx4b, GSTR, GSTP1, GSTP2, GSTO1 and GSTO2) and evaluated the Nrf2 nuclear translocation level associated with nuclear Nrf2 protein levels and cytosolic Nrf2 protein levels as well as the mRNA levels of signalling molecules (Keap1a, Keap1b and Nrf2) in fish intestines. It has been generally accepted that Nrf2 nuclear translocation could activate the Nrf2 signalling pathway to regulate the genes expressions of antioxidant enzymes in mammals 58 . Our results innovatively found that magnesium deficiency may suppress the nuclear translocation of Nrf2 to inhibit the Nrf2 signalling pathway and downregulate the expression of genes corresponding to antioxidant enzymes, which then decreases the antioxidant enzyme activities in fish intestines. (2) The caspase family proteins (such as caspase-2, -3, -7, -8 and -9) are mainly in charge of cell apoptosis in organisms 8 . In the caspase family, apoptosis executioner (such as caspase-3 and -7) can directly regulate cell apoptosis in organisms 8 . Apart from the apoptosis executioners (such as caspase-3 and -7), the apoptosis activators (such as caspase-2, -8 and -9) of the caspase family are the upstream signalling molecules of apoptosis executioners (such as caspase-3 and -7), which can directly activate the apoptosis executioners (such as caspase-3 and -7) to regulate cell apoptosis in organisms 8 . Additionally, there are two distinctly-different apoptotic pathways (the death receptor pathway and the mitochondria pathway) in organisms, and the pathways include several signalling molecules [pro-apoptotic proteins (FasL, Apaf-1 and Bax) as well as anti-apoptotic proteins (Bcl-2, Mcl-1b and IAP)] to regulate the caspases (caspase-2, -3, -7, -8 and -9) taking part in cell apoptosis 92 . However, previous researches involving the effect of magnesium on cell apoptosis in animals has only focused on one apoptosis executioner (caspase-3 activity), a protein which is part of the terminal apoptosis signalling pathway 11,93 . Surprisingly, in our present study, we innovatively found that magnesium deficiency could upreguate the gene expressions of caspase-2, -3, -7, -8, -9 and pro-apoptotic proteins (FasL, Apaf-1 and Bax), but downregulate the gene expressions of anti-apoptotic proteins (Bcl-2, Mcl-1b and IAP) to activate the death receptor pathway [FasL/caspase-8/(caspase-3 and -7)] and the mitochondria pathway [(Bax, Apaf-1, Bcl-2, Mcl-1b and IAP)/(caspase-2 and -9)]/(caspase-3 and -7)], aggravating cell apoptosis in fish intestines. (3) It is generally accepted that scaffolding proteins (such as ZO-1 and ZO-2), barrier-forming proteins (such as claudin-c, -3, -f, -b, -11, -7 and occludin) and pore-forming TJs (such as claudin-12 and -15) of tight junctional complexes (TJs) play different roles in corporately maintaining the animal intercellular integrity [94][95][96] . Additionally, MLCK can induce contraction of the perijunctional actomyosin ring that encircles the cell at the adherens junction and TJ through myosin II regulatory light chain phosphorylation to regulate the tight junction permeability in terrestrial animals 79 . Until now, only one previous study involving the effect of magnesium on TJs in animals focused only on one of the scaffolding proteins (ZO-1) and one of the barrier-forming proteins (occludin) 5 . However, our present study observed that magnesium deficiency could downregulate the mRNA levels of scaffolding proteins (ZO-1), barrier-forming proteins (claudin-c, -3c, -f, -b, and -11 and occludin) and pore-forming TJs (claudin-12), but upregulate the mRNA levels of scaffolding proteins ZO-2b (only in PI), pore-forming TJs (claudin-15a and SCIENtIFIC RePoRTS | (2018) 8:12705 | DOI:10.1038/s41598-018-30485-8 -15b) and MLCK, indicating that magnesium deficiency could activate the MLCK signalling pathway to regulate the tight junctional complex (TJs) function in animals.
In the present study, we investigated the magnesium requirement of grass carp (223.85-757.33) g based on growth performance. Based on the PWG of grass carp (223.85-757.33 g), the optimal dietary magnesium level was calculated to be 770.38 mg/kg (Fig. 7), which is slightly higher than that of juvenile grass carp (7.69-35.90 g), for which 713.50 mg/kg was recommended by Wang et al. 41 . The reason for the different optimal dietary magnesium levels between grass carp (223.85-757.33 g) and juvenile grass carp (7.69-35.90 g) may be attributed to the different growth rates of juvenile grass carp (7.69-35.90 g) and grass carp (223.85-757.33 g). As is known, due to a higher growth rate, the nutrient requirements (such as available phosphorus) for juvenile fish are higher than those for young fish 25,97 . Interestingly, however, we found that the juvenile grass carp growth rate (2.01%/day) recommended by Wang et al. 41 is much lower than the normal juvenile grass carp growth rate (3.07%/day) recommended by Dong et al. 98 . Additionally, the juvenile grass carp (7.69-35.90 g) growth rate (2.01%/day) recommended by Wang et al. 41 is close to the grass carp (223.85-757.33 g) growth rate (2.03%/day) in our study, which resulted in the slightly higher magnesium requirement for grass carp (223.85-757.33 g).
Simultaneously, we also investigated the magnesium requirement of grass carp (223.85-757.33 g) based on intestinal structural integrity related indices. Based on against ROS, against caspase-9 and claudin-3c in grass carp (223.85-757.33 g) (Fig. 7), the optimal magnesium levels were calculated to be 839.86, 856.79 and 811.49 mg/ kg, respectively. Comparatively, the optimal magnesium levels based on intestinal structural integrity related indices were higher than those based on growth performance (PWG), indicating that more magnesium might be needed to maintain intestinal structural integrity in fish. This can be attributed to the additional requirements of antimicrobial-related enzymes (acid phosphatase and AKP) in fish intestines. As our data shows, pathogen invasion (such as A. hydrophila) could impair fish intestinal structural integrity. Additionally, it has been reported that magnesium could enhance the activities of AKP in calf intestines 99 and of acid phosphatase in carp (Cyprinus carpio L.) 100 . AKP could detoxify lipopolysaccharides and prevent pathogens in Zebrafish (Danio rerio) 101 , and acid phosphatase can be a marker of the digestive capacity of phagocytes, which are associated with the elimination of pathogens in fish 102 . Therefore, we assumed that when pathogens invaded, fish would need more magnesium to enhance the antimicrobial-related enzyme activities and maintain intestinal structural integrity.

Conclusion
Looking back to our research (Fig. 8), magnesium deficiency suppressed fish growth, and we systematically investigated the impacts of magnesium deficiency on structural integrity in fish intestines. The following novel results were discovered in this study. (1) Magnesium deficiency weakened the antioxidant ability to impair the cellular structural integrity, which was attributed to the suppression of Nrf2 nuclear translocation that inhibited the Nrf2  Experimental designs for diets. Feed contents and nutrient levels are presented in Table 4. Magnesium sulfate (MgSO 4 ·H 2 O) was supplemented at 0 (control diet), 200, 400, 600, 800 and 1000 mg/kg in the basal diet. According to atomic absorption spectrometry 103  Experimental process and sample collection. Fish in this study were got from the fisheries of Sichuan, China. Prior to the growth trial, the grass carp were supplied with the control diet for 14 days according to Wang et al. 41 . After a 14 days acclimatization period, the total 540 grass carp [approximately 223.85 (SD 0.49) g] were randomly allocated into six treatments and each treatment has triplicate cages (30 fish/cage in each cage). And each cage (1.4 L × 1.4 W × 1.4 H m) was furnished with a round plate (diameter 100.00 cm) at the bottom, aiming to collect the residual feed. During growth trial, grass carp were supplied with their corresponding diet to apparent satiety four times per day. Any uneaten feed were collected thirty minutes later after feeding, then the feed not eaten were dried and weighed to calculate the feed intake (FI). In the experiment period, water magnesium concentration was determined to be 7.42 (SD 2.58) mg/L. The pH and water temperature were determined to be 7.5 (SD 0.3) and 28.5 (SD 2) °C, respectively. Also, oxygen content in water was maintained above 6.0 mg/L. Before and after growth trial, all grass carp in each treatment were weighed for calculating the growth performance related parameters. Later, all fish were anaesthetized with benzocaine before sacrificed. Then, the fish intestines were quickly separated and segmented into proximal intestine (PI), mid intestine (MI) and distal intestine (DI). After that, fish intestines were weighed and measured to calculate the intestinal related indices, and to measure the magnesium concentrations as well as the AKP and Na + , K + -ATPase activities. The grass carp blood samples in six treatments were obtained from fish bodies. Subsequently, the blood samples of grass carp were separated to get the serum which was saved in the −80 °C for measuring serum magnesium concentration. After the growth trial, using the prevalent pathogens to impair the fish intestinal structural integrity is a common approach to evaluate the nutritional protection on fish intestinal structural integrity 104 . The A. hydrophila is a popular pathogen which could impair fish intestinal structural integrity 83 . After a 60 days growth trial, fifteen grass carp in similar body weight were selected from each treatment group to inject the A. hydrophila into fish bodies intraperitoneally for 14 days challenge trial (Fig. S1). Besides, another fifteen grass carp were selected from the control diet group to inject physiological saline into fish bodies intraperitoneally. After the challenge trial, fish were anaesthetized with benzocaine before sacrificed. Subsequently, all fish intestines were quickly separated and segmented into three segments. And the samplings were saved in the −80 °C until analysis.

Materials and Methods
Sample preparation and biochemical parameters analysis. The intestines samples were homogenized in 10 volumes (w/v) of ice-cold physiological saline to get the homogenate. After that, the homogenate was centrifuged at 6000 g for 20 min at 4 °C to collect the supernatant which was saved for subsequent analysis of related parameters. The malondialdehyde (MDA), ROS, glutathione (GSH) and protein carbonyl (PC) contents were determined according to previous studies 105,106 . The anti-hydroxy radical (AHR) and anti-superoxide anion (ASA) capacities were determined according to Feng et al. 107 . Besides, the copper, zinc superoxide dismutase (CuZnSOD), total superoxide dismutase (SOD), catalase (CAT), glutathione-S-transferases (GST) and glutathione peroxidase (GPx) activities were determined as described by pervious studies 108,109 . The activity of glutathione reductase (GR) was measured according to Yang et al. 110 . Additionally, the total SOD activity minus CuZnSOD activity to get the manganese superoxide dismutase (MnSOD) activity. The analytical methods of the magnesium concentration in serum and in grass carp intestines are similar to Wang et al. 41 . The intestinal alkaline phosphatase (AKP) and NA + -K + -ATPase activities can be measured according to previous study 111 . Histological changes. Intestinal histological samples were rinsed in saline and preserved in 4% paraformaldehyde solution. Subsequently, the preserved intestinal samples were clear and dehydrated in a series of increasing ethanol concentrations (70%, 80%, 85%, 90%, 95% and 100%). After that, the tissues were prepared for being embedded in paraffin wax and sectioned to 4 mm. And sections were prepared for using standard hematoxylin and eosin (H & E) to be stained as described by Wang et al. 112 . After stained, the histological sections were examined by using a Nikon TS100 light microscope.
Detection of fragmentation in DNA. The DNA fragmentation in different intestinal segments was isolated with reference to Kawakami et al. 113 . Fragmented DNA was assayed by agarose gel electrophoresis. The DNA was loaded on to the 2.0% agarose gel, and then electrophoresis was carried out at 80 V for 1.5 h. The gel was visualized and photographed by the Gene Genius Bio-Imaging system (Syngene, Frederick, MD, USA).  Analysis of real-time PCR. The total RNA of fish intestines was isolated with RNAiso Plus kit (TaKaRa, Dalian, Liaoning, China). Single-stranded cDNA was prepared from total RNA by using the PrimeScript ™ RT reagent Kit (TaKaRa) to reverse transcription. PCR Specific primers were designed based on gene sequences that were publicly available at gene bank of grass carp and cloned by our laboratory for quantitative real-time PCR (Table 5). By evaluating the internal control genes in our preliminary experiment (Tables S1-S3), β-actin was used as a reference gene to normalize cDNA loading. According to the specific gene standard curves, the housekeeping and target gene amplification efficiency were calculated. Besides, the amplification efficiency of primers are listed in Table 5. According to Schmittgen et al. 114 , the calculation method in expression results was the 2 −ΔΔCT method. Western blotting. Protein homogenates were prepared from intestines. Antibodies incubation and western blot analysis were processed according to Jiang et al. 115 . When the intestinal protein was extracted, the concentrations of protein were measured by using corresponding assay kit. After being separated by SDS-PAGE, protein samples (40 μg per lane) were transferred to a PVDF membrane for analysis of western blot. After being blocked at room temperature for 1.5 h, then the membrane was incubated overnight at 4 °C with primary antibody. The nuclear Nrf2, cytosolic Nrf2, Lamin B1and β-Actin antibodies were the same with those in previous study 115 . In this study, nuclear Nrf2 employed Lamin B1 as control proteins and cytosolic Nrf2 used β-Actin for control proteins. Subsequently, the PVDF membrane got washed with TBST for three times with 5 minutes each time and got incubated with goat anti-rabbit horseradish peroxidase-conjugated secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA) in TBST for 2 h. The immunoreactive bands were visualized by ECL reagents (Beyotime Biotechnology Inc., China). The density of protein bands were detected by using NIH Image 1.63 software. Results for all protein levels by densitometric analyses were expressed as the fold of nucleotides treatment groups relative to the control group. The western blotting result from each group were measured there times independently.
Data statistics and analysis. Growth performance parameters were calculated, and the formulas were The analysis method of data was the one-way analysis of variance (ANOVA), and the significant differences between the treatments were calculated by Duncan's multiple-range test with SPSS 19.0 (SPSS Inc., Chicago, IL, USA) at P < 0.05. Based on PWG and structural integrity related indicators, the optimal dietary magnesium levels were calculated by quadratic regression model as described by Wang et al. 41 .