Transthyretin Upregulates Long Non-Coding RNA MEG3 by Affecting PABPC1 in Diabetic Retinopathy

The aim of the study was to demonstrate how transthyretin (TTR) could affect long non-coding RNA (lncRNA) of maternally expressed gene 3 (MEG3) and play important roles in diabetic retinopathy (DR). A DR model in C57BL/6 mice was established after intraperitoneal injection of streptozotocin (STZ). After intravitreal injection with TTR pAAV vector, MEG3 short hairpin RNA (shRNA), scrambled shRNA, or MEG3, retinal imaging, retinal trypsin digestion, and fundus vascular permeability tests were performed. Cell counting kit-8 (CCK8), transwell, and Matrigel assays were employed to detect the proliferation and migration of human retinal microvascular endothelial cells (hRECs). The binding between long non-coding RNA of maternally expressed gene 3 (lncRNA-MEG3) and microRNA-223-3p (miR-223-3p) was observed by using luciferase reporter assays, while co-immunoprecipitation (co-IP) was employed to confirm the interaction between TTR and the target. In the DR mice model, retinal vascular leakage and angiogenesis were repressed by overexpressing TTR. In vitro, the added TTR promoted the level of lncRNA-MEG3 by interacting with poly (A) binding protein cytoplasmic 1 (PABPC1), and then repressed proliferation and angiogenesis of hRECs. In vivo, silencing or overexpressing lncRNA-MEG3 significantly affected retinal vascular phenotypes. Additionally, the interaction between lncRNA-MEG3 and miR-223-3p was confirmed, and silencing of miR-223-3p revealed similar effects on hRECs as overexpression of lncRNA-MEG3. In summary, in the DR environment, TTR might affect the lncRNA MEG3/miR-223-3p axis by the direct binding with PABPC1, and finally repress retinal vessel proliferation.


Introduction
Diabetic retinopathy (DR) is considered to be among the most severe causes of vision impairment and loss in the working-aged and elderly population [1], and the incidence of DR increases annually worldwide [2][3][4][5]. Due to the complex etiology of DR and other factors, the pathogenesis of DR is not entirely clear [6]. Therefore, continuing research is necessary to elucidate the pathogenesis and underlying molecular mechanisms of the development and progression of DR.

The Protective Effects of TTR on Retinas of DR Mice
The progression of DR was characterized by abnormal retinal microvasculature, reduced retinal perfusion, increased vascular permeability, and pathological intraocular proliferation of retinal vessels. Diabetic mice were induced with intraperitoneal injection of streptozotocin (STZ), and after intravitreal injection with TTR pAAV vector, MEG3 short hairpin RNA (shRNA), scrambled shRNA, or MEG3, the serum glucose level and weight of these mice revealed no significant fluctuations after three months ( Table 1). As evaluated by the Evans Blue (EB) leakage assay, overexpressed TTR repressed diabetes-associated retinal vascular leakage ( Figure 1A,B), and the retinal trypsin digestion assay demonstrated that overexpressed TTR partially reversed vascular hyperglycemia-induced pericyte loss and aggravated capillary degeneration ( Figure 1C,D). In addition, the results of the retinal optical tomography scan and fundus photography assays exhibited significantly reduced hard exudate in diabetic mice. Reduced overexpression of TTR resulted in changes that could be observed in DR, partially reversing vascular distortion ( Figure 1E,F). , and STZ + TTR (n = 4) mice infused with Evans blue dye for 2 h were stripped and scanned using a fluorescence microscope. The severe retinal vascular leakage of STZ-induced diabetic retinopathy (DR) mice could be reversed with the overexpressed TTR (** p < 0.01) (×100 magnification times). (C,D) The retinas of control (n = 6), STZ (n = 3), and STZ + TTR (n = 4) mice were applied for trypsin digestion; the acellular retinas of STZ-induced DR mice revealed a significantly higher capillary number (quantified in 10 fields per retina, pointed with red arrows), while overexpressed TTR showed the reverse of this phenomenon (** p < 0.01) (×400 magnification times). (E) Spectral domain fundal photographs of normal, STZ, and TTR mice; there was no significant difference (×100 magnification times). (F) Spectral domain optical coherence tomography (OCT) imaging of normal, STZ, and TTR mice, and the hard exudate of the retina is marked with red arrows (×400 magnification times). All data are shown as mean ± SEM. The difference in physiological parameters between non-diabetic and diabetic mice was determined by repeated measures ANOVA at different time points. "*" indicates a significant difference compared with the Ctrl group. scr: scrambled short-hairpin RNA.

TTR Interacted with Poly(A) Binding Protein Cytoplasmic 1 (PABPC1) to Stabilize lncRNA-MEG3
Co-immunoprecipitation (Co-IP) assays in human embryonic kidney 293 (HEK293) cells confirmed the interaction between TTR and PABPC1 ( Figure 2A); in hRECs, western blot analysis indicated that PABPC1 levels increased with TTR, while lncRNA-MEG3 overexpression or knockdown showed no significant effects on the PABPC1 level with TTR ( Figure 2B) in high-glucose environments. Previous studies demonstrated that PABPC1 participates in regulation pathways and lncRNA stabilization [33][34][35]. To investigate whether PABPC1 regulated the stability of lncRNA-MEG3, actinomycin D was used to inhibit RNA synthesis in hRECs. PABPC1 knockdown significantly and rapidly decreased the level of lncRNA-MEG3, while overexpression of PABPC1 caused the reverse ( Figure 2C). These results indicated that TTR enhances PABPC1 content by interacting with PABPC1 directly, and could even further stabilize lncRNA-MEG3 under in vitro hyperglycemic conditions. was confirmed by co-immunoprecipitation (co-IP). PABPC1 and TTR were expressed or co-expressed in HEK293 cells; PABPC1 and TTR were fused with Flag-and GFP-tags, respectively. B indicates blank (IgG), -P indicates PABPC1, -T indicates TTR, and -P + T represents co-expressed PABPC1 and TTR. Input samples were used as controls without Protein A beads. Protein A beads-Anti Flag IgG was used to capture PABPC1 or PABPC1-TTR complex, while Protein A beads-Anti GFP IgG was used to capture TTR or the TTR-PABPC1 complex. (B) The levels of PABPC1 was tested by western blot in HG, TTR, and TTR with overexpression of lncRNA-MEG3 (OE lncRNA-MEG3) or small interfering lncRNA-MEG3 (silncRNA-MEG3). (C) The relationship between lncRNA-MEG3 and PABPC1 was tested by qRT-PCR and this experiment was repeated three times (* p < 0.05).

TTR Effects Proliferation, Migration, and Vascularization by Regulating lncRNA-MEG3
Our previous work has exhibited that TTR inhibits retinal vascular proliferation [14][15][16][17]. In this current work, lncRNA-MEG3 levels were increased by adding TTR ( Figure 3A). After hRECs were treated with pcDNA3.1-OEMEG3, pcDNA3.1-siMEG3, pcDNA3.1-siMEG3, and TTR, or treated with pcDNA3.1-sicontrol and TTR, cell counting kit-8 (CCK8) analysis revealed that lncRNA-MEG3 overexpression inhibited hREC proliferation, while lncRNA-MEG3 knockdown reversed TTR inhibition of hREC proliferation ( Figure 3B). The wound healing, transwell, and tube formation assays demonstrated that during lncRNA-MEG3 knockdown, TTR did not significantly affect hREC proliferation, migration, and vascularization, and lncRNA-MEG3 overexpression enhanced TTR inhibition of hRECs ( Figure 3C-E). In wound healing analysis, overexpression of MEG3 or the addition of exogenous TTR could significantly repress the wound healing process, while knockdown of MEG3 showed the opposite phenomenon (repeated independently three times; *, p < 0.05) (×100 magnification times). (D) In transwell assay, overexpression of MEG3 or the addition of exogenous TTR could significantly repress cell migration, while knockdown of MEG3 enhanced the migration trend (*, p < 0.05) (×100 magnification times). (E) In tube formation assay, tube-like structures were observed 6 h after cell seeding. Average length of tube formation for each field was statistically analyzed for at least three independent experiments. Overexpression of MEG3 or the addition of exogenous TTR could significantly suppress tube formation process of hRECs, while knockdown of MEG3 enhanced it (*, p < 0.05) (×100 magnification times). (F) In western blot analysis, TTR could regulate the levels of key proteins in Notch pathway; overexpression of MEG3 could enhance this trend, while knockdown of MEG3 could partially rescue it.
The Notch signal pathway plays vital roles in cell proliferation [36], and miR-223-3p has been proved to promote the level of Notch1 in hRECs in our previous work [32]. In order to find out whether the level of MEG3 could act on this pathway, western blot analysis was employed; the results showed that TTR treatment suppressed the levels of key factors in this pathway including Notch1, c-Notch1 (cleaved Notch1), Jagged-1, Delat-like protein 4 (DLL4), and Hes-1; lncRNA-MEG3 overexpression enhanced this trend, while lncRNA-MEG3 knockdown partially rescued this phenomenon ( Figure 3F). These data suggest that TTR could inhibit neovascularization by upregulating the level of lncRNA-MEG3.
In vivo, overexpression of lncRNA-MEG3 educed retinal hard exudates, cellular capillary number, and retinal vascular leakage, while lncRNA-MEG3 shRNA had the opposite effects. Further, TTR reversed the retinal dysfunction resulting from lncRNA-MEG3 shRNA ( Figure 4A-C). These data indicate that TTR might affect DR neovascularization through lncRNA-MEG3. , and STZ + MEG3 shRNA + TTR (n = 4) mice were applied for trypsin digestion. Overexpression of MEG3 and TTR could reduce the capillary number in acellular retinas of STZ-induced DR mice (quantified in 30 fields per retina and average, signaled with red arrows), and knockdown of MEG3 showed an opposite function, using an empty vector and scr as blanks (*, p < 0.05; **, p < 0.01) (×400 magnification times). (C) Spectral domain fundal photographs showed no significant difference. In spectral domain optical coherence tomography (OCT) imaging of mice, the hard exudate of retina was marked with red arrows; overexpression of MEG3 and TTR could reduce hard exudate of retina, but knockdown of MEG3 enhanced it. scr: scrambled short-hairpin RNA.
In this study, "RegRNA" and "MiRCode" were used to predict potential targets of miR-223-3p and lncRNA-MEG3 was indicated as a probable candidate with two miR-223-3p binding sites ( Figure 5A). In dual luciferase reporter gene assays, wild-type lncRNA-MEG3 and mutant sequences on one or both of the two binding sites (Site 1, Site 2, or Site 1 with Site 2) were employed in a luciferase vector. miR-223-3p significantly reduced luciferase activity to 42% for wild type lncRNA-MEG3, and miR-223-3p suppressed luciferase activities to 62% and 89% for mutant lncRNA-MEG3 (Site 1 or Site 2, respectively). No shift of luciferase activities for miR-223-3p was observed on the double mutation (Site 1 with Site 2) of lncRNA-MEG3 ( Figure 5B).
We compared the abundance of lncRNA-MEG3 than that of miR-223-3p to serve as a ceRNA. The levels of lncRNA-MEG3 and miR-223-3pwere calculated as 1002 and 1100 copies per cell, respectively ( Figure 5C).

Overexpression of miR-223-3p Rescues the Inhibitory Effect of TTR on Neovascularization
After the knockdown and overexpression of endogenous miR-223-3p, the level of lncRNA-MEG3 did not significantly fluctuate, which suggested miR-223-3p should be the downstream target of lncRNA-MEG3 ( Figure 6A). It is noticeable that overexpression of miR-223-3p significantly promoted neovascularization, while the addition of TTR reversed this phenomenon ( Figure 6B-E). In CCK8 assay, the miR-223-3p mimic could promote the proliferation hRECs, while TTR could rescue partially rescue this trend (repeated independently three times) (C) In wound healing assay, the miR-223-3p mimic could increase the healing process, but TTR could repress the wound healing of hRECs (repeated independently three times; *, p < 0.05) (×100 magnification times). (D) In tranwell assay, miR-223-3p mimic could enhance the migration of hRECs, while TTR showed an inhibition function (repeated independently three times) (×100 magnification times). (E) In tube formation assay of hRECs, miR-223-3p mimic promoted vascularization; however, TTR repressed angiopoiesis (repeated independently three times; *, p < 0.05) (×100 magnification times). (F) The effects of miR-223-3p and TTR on Notch signal pathway was investigated using western blot analysis.

TTR/LncRNA-MEG3/Axis Regulated Neovascularization in DR
To elucidate the role of the TTR/LncRNA-MEG3/miR-223-3p axis, it was shown that TTR could significantly reverse high level of miR 223-3p induced by MEG3 knockdown, while lncRNA-MEG3 overexpression reduced miR-223-3p levels, and it was in accordance with TTR. TTR with siMEG3 rescued the diminished miR-223-3p levels ( Figure 7A). Western blot assays showed that TTR and lncRNA-MEG3 overexpression downregulated the levels of Notch1, c-Notch1, Jagged-1, DLL4, and Hes-1, while miR-223-3p mimic or inhibitor partially rescued or promoted the above phenomenon ( Figure 7B,C). These results have suggested that TTR should inhibit the neovascularization process in DR by regulating theMEG3/miR-223-3p axis. (A) In qRT-PCR analysis, TTR or overexpression or knockdown of MEG3 could significantly regulate the level of miR-223-3p (repeated independently three times; *, p < 0.05, **, p < 0.01). (B,C) Using western blot assay, miR-223-3p mimic could enhance the Notch pathway and then could further promote angiopoiesis; however, high abundance of TTR could reverse the function of miR-223-3p by regulating the level of lncRNA MEG3, which had been proved to be the ceRNA of miR-223-3p. The experiment was repeated independently three times.

Discussion
DR is associated with visual loss, and therefore a comprehensive understanding of the underlying mechanisms of DR might identify useful diagnostic or therapeutic targets [39]. Regarding some clinical investigations of ocular diseases, myopia is suggested to prevent patients from suffering DR [11,12]; therefore, in our previous work, we studied the relationship between the higher vitreous TTR level of high myopia patients [13] and the DR protection phenomenon. It is interesting that the serum and vitreous TTR levels in DR patients should be associated with DR progression [15], and in vitro, TTR has been proved to prevent the progression of DR [14], and repress angiogenesis by the Tie signaling pathway in hyperglycemia [16] or promote the apoptosis of hRECs by the GRP78-dependent pathway [17]. Additionally, TTR has also been proved to suppress the proliferation of hRECs by a TTR/STAT4/miR-223-3p/FBXW7 signaling pathway and could further affect the content of Notch1 [32]. However, these phenomenon and mechanism should only be a small part in the whole puzzle, and more investigations are necessary to reveal the detailed effects of TTR on the progression of DR.
In our previous work, although TTR could inhibit the proliferation and induce the apoptosis of hRECs in vitro [14][15][16][17]32], the effect of TTR on DR progression in vivo was unclear. Therefore, in this work, the DR mice model was induced with STZ, and after intravitreal injection of adeno-associated virus (AAV) vector containing TTR cDNA, the pathological progression of DR was partially reversed with overexpressed TTR (Figure 1). This suggested that TTR could repress the progression of DR both in vitro and in vivo.
On one hand, to find out how could TTR affect miR-223-3p, immunoprecipitation (IP) was used to screen the direct target of TTR and total bind samples were applied for LTQ-MS analysis, while PABPC1 was identified with high abundance (data not shown). Furthermore, the directed interaction between TTR and PABPC1 was confirmed by co-IP (Figure 2A).
On another hand, to search for the downstream factor of miR-223-3p was a challenge; as reported, lncRNA could regulate biological functions as a ceRNA or molecular "sponge" for other RNA molecules [40,41]. Several lncRNAs associated with ocular diseases and vascular progression were considered as potential candidates, including MEG3 [30,31]. In this work, the direct binding of lncRNA-MEG3 and miR-223 at two sites was proved in luciferase reporter assays ( Figure 5); overexpression or knockdown of lncRNA-MEG3 could significantly decrease or enhance the level of miR-223-3p ( Figure 7A), but miR-223-3p could not reveal the same effects on lncRNA-MEG3 ( Figure 6A), which suggested that miR-223 should be the direct downstream target of lncRNA-MEG3. In vitro, addition of TTR could enhance the level of lncRNA-MEG3, and overexpression of lncRNA-MEG3 revealed the same repression function on the proliferation, migration, and tube formation properties of hRECs as TTR (Figure 3) in hRECs. In vivo, by intravitreal injection, lncRNA-MEG3 also showed the same protection function on the pathological progression of DR in the STZ-induced DR mice model (Figure 4). These results suggested that there might be a potential relationship between TTR and MEG3.
Additionally, according to previous reports, PABPC1 could bind the poly(A) tails of mRNAs and further regulate the stability and biofunction of lncRNAs [33][34][35], which indicated a potential relationship between PABPC1 and lncRNA-MEG3. Because lncRNA-MEG3 showed no significant effects on PABPC ( Figure 2B), while the knockdown and overexpression of PABPC1 could significantly decrease or promote the level of lncRNA-MEG3 ( Figure 2C), lncRNA-MEG3 should be the downstream target or PABPC1.
Then, a TTR/PABPC1/lncRNA-MEG3/miR-223-3p signal axis was suggested by the results of this work. In our previous study, miR-223-3p has been proved to directly interact with 3' UTR of FBXW7, and further affect the Notch1 pathway [32]. Combined with the results of this work, a more complete pathway was suggested: TTR could bind to PABPC1 and increase its content, PABPC1 could further bind and stabilize lncRNA-MEG3, which could decrease the level of miR-223 and further promote the level of FBXW7, and then the Notch1 pathway could be inhibited (Figure 8).

Vascular Tube Formation Assay
A basement membrane matrix (BD Biosciences, Franklin Lakes, NJ, USA) was used to detect tube formation. Matrix was loaded in a 48-well plate, and after drying for 3 min at 37 • C, 3 × 10 4 hRECs were seeded in each well, followed by 6 h of incubation at 37 • C. Vascular tube formation of the hRECs was evaluated using an Olympus microscope (IX-73).

Cell Migration and Healing Assay
To assess cell migration, hRECs were seeded into upper wells and incubated 24 h, then hRECs were fixed on the inserts with 4% paraformaldehyde, stained with crystal violet, and counted under a light microscope. For healing assays, hRECs were incubated until confluence on a 6-well dish, then scratched with a sterile 10 µL pipette tip. After 24 h of incubation, wells were scanned with IncuCyte ZOOM (ESSEN BioScience, Ann Arbor, MI, USA), and relative wound density was evaluated.

Diabetes and DR Mouse Models Induced with Streptozotocin (STZ)
To establish models of diabetes and DR in mice, 50 µg/g of STZ were injected intraperitoneally daily for 5 d into 8-week old male C57BL/6 mice, using citrate buffer as a blank. The mice were considered to model diabetes once blood glucose surpassed 16.7 mmol/L. After intraperitoneal injection of 80 µg/g ketamine and 4 µg/g xylazine, mice were ventilated, and approximately 1.5 µL (1×10 12 TU/mL) of adeno-associated virus containing TTR cDNA were intravitreally injected. For short-hairpin RNA (shRNA) or MEG3 shRNA adenovirus, mice were intravitreally injected every three weeks.
There were two main animal treated assays designed (Evans blue leakage and Retina trypsing digestion assays), and after these treatments, mice could not be used repeatedly. Therefore, mice were divided into two groups for testing. Details of animal amounts have been exhibited in Table 1 and Figures 1 and 4.

Retinal Imaging
After mice were anesthetized with 80 µg/g ketamine and 4 µg/g xylazine, Cyclometric (Alcon, Fort Worth, TX, USA) was used to dilate the pupils. A Micron IV image-guided OCT system (Phoenix Research Laboratories, Pleasanton, CA, USA) was then employed to detect spectral domain optical coherence tomography, guided by bright-field live fundus imaging.

Evans Blue (EB) Leak Assay
After mice were anesthetized with 80 µg/g ketamine and 4 µg/g xylazine, the mice were cannulated at the right jugular vein and the iliac artery, injected with heparinized saline, and injected with 45 µg/g EB into the jugular vein. Approximately 0.2 mL of blood were extracted from the mice after 2 h, then mice were perfused with PBS followed by 1% paraformaldehyde to the left ventricle. After the cornea, lens, and vitreous humor were depleted, the retina and sclera were fixed in 4% paraformaldehyde at 25 • C for 3 min. The retina was incubated in dimethylformamide at 78 • C for 12 h and centrifuged for 15 min at 12000× g before detection at A 620 nm (blue) and A 740 nm (background).

Trypsin Digestion of Retina
After enucleation, mice eyeballs were fixed in 4% paraformaldehyde at 25 • C for 24 h. Retinas were extracted and digested with 3% trypsin at 37 • C for 3 h; the retinal vasculature was then washed, located on glass slides, dried, and stained with periodic acid-Schiff.

Luciferase Reporter Assays
Transient transfections were performed using the Lipofectamine2000 transfection system (Invitrogen) on human embryonic kidney 293 (HEK293) 50,000 cells plated in a 24-well plate. At 24 h after transfection, the cells were lysed in 100 µL of passive lysis buffer, and firefly and Renilla activity was determined with a luminometer using the Dual Luciferase Assay System (Promega, Madison, WI) on 20 µL of lysate following the manufacturer's handbook. The pGL3-luc plasmids containing luciferase reporter gene and MEG3 binding sites or wild/mutated miR-223-3p mimic were ordered from GenePharma (Shanghai, China).

Statistical Analysis
Data are presented as the mean ± SD from at least 3-5 independent experiments. Statistical comparisons between two groups were made using two-tailed Student's t tests. Differences among three or more groups were compared by one-way ANOVA followed by least significant difference post hoc tests (SPSS 17.0, Chicago, IL, USA). p < 0.05 was considered as statistically significant.

Conclusions
In summary, this work demonstrated that TTR might have a critical role in the progression of DR, that it directly binds PABPC1, and that it stabilizes lncRNA-MEG3. Additionally, lncRNA-MEG3 might inhibit DR neovascularization by affecting the Notch pathway, and the TTR/lncRNA-MEG3/miR-223-3p axis mediates anti-angiogenesis in DR. These results provide novel insights on the molecular pathogenesis of DR, and will inform future research into novel lncRNA-directed diagnostic and therapeutic targets for clinical prevention and therapy of DR. Still, the exact mechanism of PABPC1 regulation of lncRNA-MEG3 remains unknown and merits further research. The present findings provide a new perspective to understand the relationships between lncRNAs and their interactions with proteins.

Conflicts of Interest:
The authors declare no conflict of interest.