Thapsigargin (THAP, T9033) was purchased from Sigma–Aldrich Company with a purity >98% (HPLC). H89 (S1643, purity 98%) and forskolin (FSK, S1612, purity 99%) were from Biotime Company. Exendin-4 (HY-13443) was from MedChem Express Company with a purity 98.96%. RPIM1640 medium (C11875500BT), DMEM medium (C11995500BT) FBS (10099141), and supplements were from Thermo Fisher, MTT (97062-376, purity 98%) and puromycin (10mg/ml, AAJ67236-8EQ) was purchased from VWR. DAPI (KGR0001-1) was from Keygen Biotech. Polybrene Transfection Reagent (TR-1003-G) was purchased from EMD Millipore Corporation. pBiFC-VC155 (P0681) and pBiFC-VN173 (P0680) were from Miaolingbio Company.

INS-1 832/13 cells (INS-1 cells) were a gift from Professor Yong Liu and grown in RPMI 1640 medium as previously reported (Yao et al., 2015; Lu et al., 2016a). All experiments were performed with this culture unless otherwise specified.

PKA Cα (Gene ID: 25636) was inserted into both pEYFP-C1 vector at SalI and BamHI site and pBiFC-VC155 vector at EcoRI and KpnI site. TXNIP (Gene ID: 117514) was inserted into both pECFP-N1 vector at NheI and HindIII site and pcDNA3.1(+)-Myc-His A vector at HindIII and BamHI site. TXNIP (S308A) or TXNIP (S307/308A) mutated genes were inserted into pBiFC-VN173 vector at HindIII and XbaI site and pcDNA3.1(+)-Myc-His A vector at HindIII and BamHI site. Lipo2000 (ThermoFisher, CA, USA) were used for transient transfections for western blotting, confocal and Co-IP assay.

INS-1 cells were cultured in 96-well plates with normal cell culture overnight at a density of 2 × 10⁴ cells/well. These cells were then treated with FSK or exendin-4 and THAP (0.5 µM) for 24 h, following 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, 0.5 mg/ml) incubation for 4 h. DMSO (200 µL/well) was added into the well to resolve the formazan crystals and the absorbance was detected using a Tecan microplate reader at 570 nm (Tecan, Mannedorf, Switzerland).

GSIS assay was performed as our previous report (Zhang et al., 2013; Yao et al., 2015). Briefly speaking, INS-1 cells were pre–incubated with KRB buffer contained 0.2% BSA for 2 h, followed by incubation in KRB buffer containing glucose (16.8 mM) with indicated compounds for 2 h. Then, the supernatant was collected and insulin concentration was detected by the Insulin High Range Kit (Cisbio, Billerica, France).

INS-1 cells were plated into 6-well plates and then incubated in 5% CO₂ for 48 h for Co-IP assay incubated with normal cell culture. IP was performed by the Classic IP kit (ThermoFisher, CA, USA). The cell (1 mg of protein) was lysed with lyse buffer (25 mM Tris, 150 mM NaCl, 1 mM EDTA, 1% NP-40, 5% glycerol, pH 7.0) and was used for IP with anti-TXNIP or anti- PKA Cα antibodies at 4°C for 12 h. Proteins were then centrifuged and pellets were washed 3 times with cold TNTE 0.1% (50 mM Tris–HCl buffer pH 7.4, 150 mM NaCl, 5 mM EDTA and 0.1% Triton X-100). Pellets were re-suspended in Laemmli buffer (100 mM Tris–HCl pH6.8, 20% glycerol, 2% SDS, 0.05% bromophenol blue and 100 mM DTT) and boiled for 5 min. The proteins were analyzed by western blotting assay.

Antibodies against TXNIP (14715), PKA Cα (5842) and β-actin (4970) were purchased from Cell Signaling Technology. For WB assay, INS-1 cells were cultured into plates (6-well), and then cell lysate was separated by SDS–PAGE and transferred into PVDF membranes. All membranes were visualized using FluorChem E (Protein simple, CA, USA) after incubation with indicated antibodies for 12 h. The image densitometry was analyzed by ImageJ 1.48 software.

Quantitative reverse transcription-PCR (qRT-PCR) experiments were performed using LightCycler480 (Roche Diagnostics International, Rotkreuz, Switzerland) using GoTaq^® qPCR Master Mix Kit (Promega Corporation, WI, USA) and PrimeScript^™RT Master Mix Kit (Takala Bio Inc, Kusatsu, Japan). Total RNA was isolated using Cell Total RNA Isolation Kit (Foregene, Sichuan, China). The concentration and the quality of RNA were measured using DS-11 FX Spectrophotometer (DeNovix Inc, DE, USA). Primers used for the qRT-PCR are given below:

The basic principle of the BiFC assay is to fuse two non-fluorescent fragments from a split fluorescent Venus protein to two interaction partners. If the two proteins interact with each other, the interaction brings the two non-fluorescent fragments together to allow reconstitution of an intact fluorescent protein molecule. Here, pBiFC-VN173 and pBiFC-VC155 were used. PKA Cα and TXNIP were fused into these plasmids separately. Then these plasmids were transfected into INS-1 cells for 24 h. After being fixed with 3.7% formaldehyde in PBS and stained with DAPI, the cells were visualized using a 100x objective lens to observe the BiFC signal (Ex/Em: 515/528 nm).

INS-1cells were transfected with the lentiCRISPRv2 [lentiCRISPRv2 puro was a gift from Brett Stringer (Addgene plasmid # 98290)] with the guide RNA sequence 5’-CCTCCCAATCCGCCGTAAGT-3’ against PKA Cα (Gene ID: 25636). These plasmids were packaged into retroviral particles by co-transfecting into 293T cells with lentiCRISPRv2- PKA Cα construct having plasmids expressing psPAX2 (psPAX2 was a gift from Didier Trono (Addgene plasmid # 12260)) and pCMV-VSV-G (pCMV-VSV-G was a gift from Bob Weinberg (Addgene plasmid # 8454)) (Stewart et al., 2003). Supernatants were collected and used to transduce 10⁵ INS1 cells in the presence of 10 µg/mL polybrene. After 24 h post transduction, cells were selected in 1 µg/mL puromycin.

Puromycin-resistant cells were single-cell sorted into a 96-well plate and colonies were grown up. Cloned INS1 cells were screened for homozygous mutations that disrupted the coding sequence of PKA Cα as follows. 100 cell colonies were used to prepare the whole cellular DNA. The region of the guide RNA was amplified using the following primers:

Amplified genomic DNA was Sanger sequenced to determine the nature of the CRISPR-CAS9-mediated genomic disruption. A cell line with confirmed homozygous disruption of PKA Cα ( Figure 4A ) was then screened out.

All of the data are expressed as mean ± SEM. Multiple comparisons were performed by one-way ANOVA analysis followed by Bonferroni post-hoc test (compare all pairs of columns), when variances were equal between groups. All statistical analyses were performed using Prism 5 Graph Pad software. In all cases, if the P value determined from a statistical test was less than 0.05, the difference was considered statistically significant. Values are presented as ns, *P < 0.05, **P < 0.01, and ***P < 0.001. All experiments were performed in triplicate. All statistical testing is exploratory, and not hypothesis-testing.

ER stress induces TXNIP expression, which in turn activates inflammasomes and finally β-cell death (Ortsater and Sjoholm, 2007; Oslowski et al., 2012). Whereas, ER stress could be alleviated by exendin-4 (GLP-1 receptor agonist) via PKA activation in pancreatic β-cells (Kim et al., 2010). In order to confirm these results, we thus treated INS-1 cells with thapsigargin (THAP), an ER stress inducer, to observe the effect of exendin-4 or FSK on β-cell viability, because exendin-4 or FSK both could activate PKA. Similar to the previous results, exendin-4 ( Figure 1A ) or FSK ( Figure 1B ) treatment could statistically significantly improve ER stress-induced β-cell death. Considering ER stress-induced inflammation is the cause of β-cell death (Oslowski et al., 2012), we evaluated the effects of FSK on IL1-β level. As shown in Figure 1C , THAP largely enhanced IL1-β transcription, which was reduced in the presence of exendin-4 or FSK. Therefore, we wanted to know whether the anti-inflammation effect was dependent on PKA. After PKA activation was inhibited by H89, a PKA inhibitor, IL1-β, was at the same level under ER stress condition with or without exendin-4 or FSK treatment. Moreover, H89 could not induce more IL-1β expression under ER stress, which excluded the possibility that the inhibition of PKA has other downstream effects that increase the IL-1β expression. The results indicated that PKA played a key role in the protective effect of exendin-4 or FSK.

Considering ER stress-induced TXNIP locates at the upstream of IL1-β, we therefore explored whether PKA activation could regulate TXNIP level under ER stress condition in β-cells. THAP statistically significantly induced TXNIP expression as early as 0.5 h post-treatment, which lasted for 8 h ( Figure 2A ). This observation was consistent with a previous report (Oslowski et al., 2012). However, FSK treatment largely decreased TXNIP protein level induced by ER Stress, as early as 0.5 h ( Figure 2B ). These results encouraged us to find out whether TXNIP transcriptional level was also inhibited by FSK. As shown in Figure 2C , FSK (10 µM) had no effect on the mRNA level of TXNIP induced by THAP after 0.5-4 h treatment, except 8 h. Furthermore, it wasfound that FSK reduced TXNIP mRNA level at 12, 24 and 48 h treatment in our lab (data not shown). From the above, these results indicated that FSK mainly promoted TXNIP degradation other than at the transcriptional level at short time incubation.

It is well known that the ubiquitin proteasome plays a key role in proteometabolism in the eukaryotic cells. We thus investigated the role of the proteasome in the progress of FSK-mediated TXNIP degradation. MG132 and bortezomib, two proteasome inhibitors, were used to inhibit the proteasome activity (Kitagaki et al., 2015), which could largely abolish the FSK-mediated TXNIP degradation ( Figure 2D ). Together, these data suggested that the FSK-mediated TXNIP degradation progress was mainly dependent on proteasome-mediated protein degradation system.

Moreover, recent studies suggested that TXNIP could be degraded after phosphorylation by AMPK or AKT (Shaked et al., 2011; Wu et al., 2013; Li et al., 2015; Waldhart et al., 2017; Zaragoza-Campillo and Moran, 2017). To clarify the involvement of the AMPK, AKT or other enzymes activated by FSK in TXNIP degradation process, several inhibitors of these kinases were used. INS-1 cells were pre-incubated with these enzyme inhibitors for 1 h, and were then treated with FSK or THAP for 8 h. MEK1/2 (U0126), CREB (ESI-09), AKT (Perifosine), PKC (Ro 31-8220 and ML-7) and AMPK (Compound C) inhibitors had little effect on the reduction of TXNIP expression by FSK ( Figure 2E ). Treatment with these inhibitors alone could not enhance TXNIP accumulation ( Figure 2F ). On the other hand, AKT inhibitor (MK-2206) reversed the effect of FSK on TXNIP degradation, which accounted for about 14%. However, H89 reversed the effect of FSK on TXNIP degradation about 71%. Indeed, H89 mainly reversed the effect of FSK on TXNIP level ( Figure 2E ). These results suggested a priority role of PKA on TXNIP level.

The above results indicated that there might be a relationship between PKA and TXNIP, but it still lacks of any direct evidence. The existence of direct interactions between these two proteins could be studied using BiFC assay, which has been widely used recently to validate protein-protein interactions (Shyu et al., 2008). Hence, we constructed TXNIP-VC155 and PKA Cα-VN173 plasmids, and these plasmids were transfected into INS-1 cells. The complementation only occurred when TXNIP interacted with PKA Cα, and was not be observed when a VC155 fragment alone was expressed with PKA Cα-VN173, indicating no spontaneous interactions occurred between the two Venus fragments. There were weak fluorescent signals in INS-1 cells with TXNIP-VC155 and PKA Cα-VN173 plasmids transfection ( Figure 3A ), which may because PKA Cα overexpression promoted TXNIP degradation and was confirmed in the flowing result ( Figure 4C ). On the other hand, MG132 was used to inhibit TXNIP degradation by PKA Cα, and there were bright fluorescent signals by TXNIP-VC155/PKA Cα-VN173 interactions, compared with the control group ( Figure 3A ). This phenomena was further confirmed by the above results that PKA Cα-mediated TXNIP degradation was dependent on the proteasome, which also demonstrated that TXNIP directly interacted with PKA Cα.

Furthermore, there was an overlap between TXNIP and PKA Cα position in the cytoplasm. ECFP-tagged TXNIP and EYFP-tagged PKA Cα were used to detect the location of these two proteins in INS-1 cells. As shown in Figure 3B , TXNIP (shown in green) and PKA Cα (shown in red) shared the same location (shown in yellow) in the cytoplasm of INS-1 cells.

Apart from these observations, the direct interactions of PKA Cα/TXNIP was further confirmed by Co-IP assay in INS-1 cells. PKA Cα was immunoprecipitated using anti- PKA Cα antibodies, and TXNIP was clearly found to be co-precipitated with PKA Cα ( Figure 3C ). Furthermore, PKA Cα was co-precipitated with TXNIP using anti-TXNIP antibodies as well ( Figure 3D ). To further confirm this phenomena, we transfected TXNIP and PKA Cα expression plasmids into HEK293T cells. As predicted, PKA Cα was markedly co-precipitated by anti-TXNIP antibodies ( Figure 3E ), and vice versa ( Figure 3F ). These data provided solid evidence that PKA Cα directly interacted with TXNIP.

The fact that PKA Cα interacted with TXNIP directly was what prompted us to study the potential regulatory effects of PKA Cα in modulating TXNIP level and pancreatic β-cell dysfunctions. Considering PKA Cα is important for PKA activity in β-cells (Goehring et al., 2007; Shibasaki et al., 2014), we constructed PKA Cα knockout (PKA Cα-KO) cell line for this purpose. Three sgRNAs were designed for PKA Cα knockout using CRISPR-cas9 system and sgRNA2 showed the best knockout efficiency ( Figure 4A ). A single cell clone of PKA Cα-KO was then screened out and sequenced, in which the PKA Cα was found to have a deletion mutation ( Figure 4B ). As expected, there was a high level of TXNIP level in PKA Cα-KO cells compared to the WT cells ( Figure 4C ). This data indicated that TXNIP degradation was blocked after PKA Cα knockout. In addition, the effect of FSK on TXNIP protein level was further evaluated in PKA Cα-KO cells, and THAP could hardly enhanced TXNIP level in PKA Cα-KO cells since cellular TXNIP was accumulated without its degradation by PKA (Figures 4C, D). Moreover, the TXNIP degradation ability of FSK was largely decreased in PKA Cα-KO cells compared with the control cells. On the other hand, H89 could not reverse the effect of FSK on TXNIP level in PKA Cα-KO β-cells ( Figure 4D ). These results were in accordance with the result presented in Figure 2E , which revealed an important effect of PKA Cα on TXNIP degradation.

It is widely reported that TXNIP overexpression has direct links with glucotoxicity, insulin secretion, β-cell inflammation and cell death (Chen et al., 2008; Reich et al., 2012). We thus evaluated the ability of glucose-stimulated insulin secretion (GSIS) in PKA Cα-KO cells. It was shown that insulin secretion was statistically significantly down regulated in PKA Cα-KO β-cells compared with WT β-cells in the presence of 16.8 mM glucose ( Figure 4E ). Moreover, insulin concentration reduced about 64.8% and 81% in WT β-cells and PKA Cα-KO β-cells under ER stress condition, respectively ( Figure 4E ). These data indicated that the TXNIP level was enhanced in PKA Cα-KO β-cells, thereby impairing insulin secretion. Furthermore, the inflammation level was evaluated in PKA Cα-KO β-cells. Compared to the WT β-cells, there were statistically significantly enhanced transcriptional levels of pro-inflammatory cytokines, such as IL-1β and IL-6 in PKA Cα-KO β-cells ( Figure 4F ).

Taken together, the above data demonstrated that TXNIP was largely enhanced in PKA Cα-KO β-cells, in which accumulation finally led to pancreatic β-cell dysfunctions.

Considering PKA Cα-knockout could increase TXNIP level and induce pancreatic β-cell dysfunctions, we thus speculated that PKA Cα overexpression could reduce TXNIP level and improve β-cell functions. At first, we transfected PKA Cα-EYFP or EYFP overexpression plasmid into WT or PKA Cα-KO β-cells. PKA Cα overexpression statistically significantly reduced the TXNIP level under normal conditions ( Figure 4C ). This result further confirmed that PKA Cα could promote TXNIP degradation, which was similar to the result shown in Figure 3A . Next, we detected the effect of PKA Cα overexpression on TXNIP level under ER stress condition. PKA Cα overexpression could reduce TXNIP level in both WT and PKA Cα-KO β-cells in the presence of THAP ( Figure 4C ). These data indicate that PKA Cα could directly decrease TXNIP level both under normal and ER stress condition.

Next, we evaluated the effect of PKA Cα overexpression on GSIS under TXNIP overexpression condition. PKA Cα overexpression could not enhance insulin secretion level under normal condition. However, PKA Cα overexpression could reverse the effect of TXNIP overexpression, which largely down regulated the insulin secretion level ( Figure 5A ). In addition, we found that the insulin concentration was reduced to almost half the level in PKA Cα-KO β-cells (1.5 ng/ml), compared to in WT β-cells (3 ng/ml), and this effect was moderately ameliorated by PKA Cα overexpression ( Figure 5B ).

We further evaluated the effects of PKA Cα on TXNIP-induced inflammation in β-cells. As expected, TXNIP overexpression has statistically significantly induced the mRNA level of the pro-inflammatory cytokines, IL-1β and IL-6 ( Figure 5C ). However, these genes induced by TXNIP expression were reduced with PKA Cα overexpression ( Figure 5C ). Moreover, the mRNA level of IL-1β and IL-6 were enhanced in PKA Cα-KO β-cells ( Figure 5D ). However, PKA Cα overexpression abolished the increasing rate of IL-1β and IL-6 mRNA in PKA Cα-KO β-cells, as expected. These data suggested that PKA Cα overexpression could reduce TXNIP level and β-cell inflammation.

Given that PKA Cα is a well-known phosphokinase, we analyzed whether TXNIP could be phosphorylated by PKA Cα, leading to its degradation. TXNIP is highly conserved among species, with a sequence similarly of 98∼99%. These sequences were analyzed using pkaPS software, a website able to predict S/T residues phosphorylated by PKA (http://mendel.imp.ac.at/sat/pkaPS/) (Neuberger et al., 2007). The predicted top3 phosphorylation sites of TXNIP based on the predicted score was shown in Figure 6A . S308 residue received the highest score, followed by S307, S314 ( Figure 6A ), as reported earlier (Shaked et al., 2011; Wu et al., 2013). Therefore, we constructed TXNIP (S308A) and TXNIP (S307/308A) to detect the degradation effects of PKA Cα. TXNIP (WT), TXNIP (S308A) or TXNIP (S307/308A) plasmids were transfected separately into β-cells with or without FSK. TXNIP (S308A) expression showed partial resistance to FSK-mediated degradation ( Figure 6B ). However, TXNIP (S307/308A) expression showed complete resistant ability to FSK-mediated degradation ( Figure 6B ). Moreover, TXNIP (S308/307A) expression could resist PKA Cα-mediated degradation with PKA Cα overexpression ( Figure 6C ). However, Ser314 mutation could not inhibit the PKA-medicated TXNIP degradation (data not shown). These results suggested that S307/S308 residue of TXNIP played a vital role in PKA Cα-mediated phosphorylation.

Furthermore, the above phenomena was confirmed by BiFC assay, TXNIP -VC155, TXNIP (S308A)-VC155, TXNIP (S307/308A)-VC155 or PKA Cα-VN173 plasmids were transfected into INS-1 cells to observe the fluoresce signals ( Figure 6D ). As expected, the fluorescent signal was brighter in TXNIP (S308/307A) overexpressed β-cells than in TXNIP (S308A) overexpressed β-cells. Even after FSK treatment, the fluorescence intensity was not decreased in TXNIP (S307/308A) overexpressed β-cells. These data indicated that TXNIP was phosphorylated by PKA Cα at Ser307 and Ser308 positions.

It was documented earlier that exendin-4 modulates TXNIP (Shao et al., 2010), but the downstream mechanisms are not clear. Also, whether exendin-4 has an effect on TXNIP-medicated β-cell inflammation is far from clear. Exendin-4 is reported to alleviate ER stress-induced β-cell dysfunctions via PKA here we investigated the role of PKA Cα/TXNIP pathway in exendin-4 mediated protective effects of β-cells. Similar to an earlier report (Shao et al., 2010), TXNIP level was statistically significantly reduced by exendin-4 treatment. However, there was no change in the level of TXNIP after exendin-4 treatment in PKA Cα-KO β-cells ( Figure 7A ). This data indicated that the effects of exendin-4 on TXNIP was dependent on PKA Cα activation.

Based on our above results using TXNIP (S307/S308) mutant in PKA Cα-mediated TXNIP phosphorylation, we transfected TXNIP (S307/308A) mutant plasmid into β-cells to study the effects of exendin-4. TXNIP protein level was statistically reduced by exendin-4 in β-cells transfected with TXNIP. Unlike TXNIP transfection, TXNIP (S307/308A) mutant has retained the high level of TXNIP with the incubation of exendin-4 ( Figure 7B ). This result further confirmed the role of TXNIP (S307/308A) in a PKA Cα-mediated phosphorylation event.

In addition, we evaluated the effects of exendin-4 on TXNIP under ER stress condition. TXNIP level was enhanced by THAP treatment, but when treated with exendin-4, ER stress-induced TXNIP level was statistically significantly reduced in WT cells ( Figure 7C ). Moreover, THAP has lost its TXNIP enhancement ability in PKA Cα-KO cells, which was similar to the results of Figure 4C . This might be ER stress researched saturated state in PKA Cα-KO cells, so THAP could not induce more TXNIP. However, the effects of exendin-4 disappeared in PKA Cα-KO β-cells under similar conditions ( Figure 7C ).

We further evaluated the effects of exendin-4 on inflammation in TXNIP or TXNIP (S307/308A) mutant overexpressed β-cells. Both TXNIP or TXNIP (S307/308A) overexpression up-regulated the IL-1β mRNA level. However, exendin-4 treatment decreased the IL-1β mRNA level in TXNIP overexpressed β-cells ( Figure 7D ), but the effect of exendin-4 was disappeared in TXNIP (S307/308A) mutant overexpressed β-cells ( Figure 7D ). These data clearly demonstrated the role of the PKA Cα/TXNIP pathway in the anti-inflammatory effects of exendin-4 ( Figure 7E ).