Destabilization of β-cell FIT2 by saturated fatty acids contribute to ER stress and diabetes

Western type diets are linked to obesity and diabetes partly because of their high saturated fatty acid (SFA) content. We found that SFAs, but not unsaturated fatty acids (USFAs), reduced the number of lipid droplets (LDs) within pancreatic β-cells. Mechanistically, SFAs but not USFAs disabled LD biogenesis by inducing palmitoylation and subsequent ERAD-C mediated degradation of LD formation protein, Fat storage-Inducing Transmembrane protein 2 (FIT2). Targeted ablation of FIT2 reduced β-cell LD numbers, lowered β-cell ATP levels, reduced Ca2+ signaling, downregulated β-cell transcription factors (RNA sequencing analysis), and exacerbated diet-induced diabetes in mice. Subsequent mass spectrometry studies revealed increased C16:0 ceramide accumulation in islets of mice lacking β-cell FIT2 under lipotoxic conditions. Inhibition of ceramide synthases ameliorated the enhanced ER stress. Overexpression of FIT2 increased number of intracellular LDs and rescued SFA-induced ER-stress and apoptosis thereby highlighting the protective role of FIT2 and LDs against β-cell lipotoxicity and diet-induced diabetes.

The role of FIT2 loss in SFA-induced β-cell apoptosis was further investigated through 210 its overexpression in MIN6 cells. MIN6 cells were transiently transfected with either an 211 expression vector encoding FIT2 (pcDNA3.1-FIT2) or with a control vector (pcDNA3.1-212 mock) prior to palmitate (or BSA) exposure. FIT2 levels increased by approximately 3-213 fold in FIT2-overexpressing (FIT2-OE) cells treated with BSA, compared to control 214 (mock cells) (Fig 4A, B). While palmitate exposure resulted in an approximately 95% 215 reduction of FIT2 in mock cells, this was partially attenuated in  (approximately 55% reduction) (Fig 4A, B). This suggests that overexpression of FIT2 217 partially compensated for palmitate-induced reduction in FIT2 protein levels. This 218 partial restoration of FIT2 in FIT-OE cells exposed to palmitate led to a modest but 219 significant increase in the number of LDs (Fig 4C, D), suggesting a partial rescue of 220 palmitate-induced LD loss in β-cells. This rescue followed with a corresponding 221 reduction in CHOP protein levels and lowered caspase-3/7 activity in FIT2-OE cells 222 exposed to palmitate (Fig 4E, F). Together, these results suggest that partial 223 compensation of FIT2 loss ameliorates palmitate-induced lipotoxicity in β-cells. 224 225 Palmitoylation and ERAD-C degradation pathways contribute to FIT2 protein 226

loss. 227
We next sought to elucidate the mechanism that may account for SFA-induced FIT2 228 loss. The observed SFA-mediated downregulation of FIT2 in MIN6 cells was not seen 229 at the transcriptional level ( Supplementary Fig 5A) but was abrogated in the presence 230 of proteasome inhibitor MG132 (Fig 5A, B), suggesting that SFAs modulate FIT2 231 protein stability rather than its gene expression in β-cells. Given that FIT2 is an ER-232 resident protein, we then probed the possible involvement of an ER-associated 233 degradation (ERAD) pathway in palmitate-mediated degradation of FIT2. ERAD 234 inhibitor Eeyarestatin I (ES) (30), partially attenuated palmitate induced FIT2 loss (Fig 235 5C,D). Increased association between FIT2 and MARCH6, a mammalian E3 ligase 236 complex responsible for degradation of the ERAD-cytosolic (ERAD-C) substrates (31), 237 was observed in MIN6 cells treated with palmitate ( Fig 5E). Furthermore, MARCH6 238 silencing (Supplementary Fig 5B) modestly but significantly rescued palmitate 239 mediated FIT2 loss (Fig 5F, G) and taken together, these results implicate ERAD-C 240 pathway and MARCH6 are involved in palmitate-mediated degradation of FIT2. The 241 ERAD-C pathway is usually triggered by a change in the tertiary conformation of an 242 ER-protein. We tested the possibility of palmitate directly modifying FIT2 given that an 243 in silico analysis (GPS-lipid) (32) showed high probability of fatty-acid S-acylation 244 taking place at 4 different cysteine residues  (Supplementary Fig 5C). Indeed, the presence of S-acylation (palmitoylation) on FIT2 246 protein was confirmed with the S-palmitoylation assay, with higher levels of S-acylated 247 FIT2 proteins detected in MIN6 cells exposed to palmitate (Fig 5H). Mutation of all 4 248 predicted S-acylation sites (Cys to Ala) abrogated palmitoylation under steady-state 249 (BSA) and heightened (palmitate) conditions ( Fig 5H). Next, to determine whether 250 FIT2 palmitoylation led to its degradation, the palmitoylation inhibitor Cerulenin (33) In a contrasting situation where LDs were allowed to accumulate in β-cells through 313 desnutrin (a TAG hydrolase) ablation, a similar blunted GSIS response was observed 314 (44). These results seem contradictory to what is reported here as they suggest that 315 the converse, intracellular LD accumulation, is detrimental to β-cells. However, it is 316 noteworthy that the significant driver for the reduced GSIS in the desnutrin knockout 317 mice is reduced fatty acid utilization by mitochondria, rather than enhanced β-cell ER 318 stress. The latter mechanism that supports our findings. To reconcile this perceived 319 difference, it is important to note that while LD formation and lipid sequestration away 320 from the ER mitigates lipotoxicity, this is perhaps on the premise that LD utilization 321 and TAG hydrolysis remains unaffected. Here, further work is required to delineate the 322 dynamics of LD accumulation and LD turnover, especially since higher number of LDs 323 has been observed in human diabetic β-cells (16). Nevertheless, it is increasingly clear 324 that the formation of LDs, especially during lipotoxic conditions, is critical for preventing 325 ER lipid accumulation and ER stress. 326 327 A fundamental mechanism governing LD loss is the post-translational modification of 328 FIT2 by palmitate (and stearate) and its subsequent degradation in β-cells. This may 329 explain the lack of FIT2 being mentioned in any GWAS or islet RNA-seq related 330 studies, although a recent diabetes meta-analysis study identified an East Asian 331 diabetes-associated loci (RS6017317) in the regulatory region of FIT2 (45). The 332 molecular switch of this protein, and a potential therapeutic strategy, likely involves 333 protein stability maintenance at the ER, as evidenced by a partial rescue of lipotoxicity 334 through FIT2 overexpression alone. Palmitoylation commonly occurs on 335 transmembrane proteins, affecting protein stability and subcellular localization (46,47). 336 Interfering with palmitoylation either pharmacologically or through site-directed 337 mutagenesis reduced palmitate-induced FIT2 degradation, and a similar mechanism 338 was reported in other transmembrane proteins such as TBC1D3 and CDCP1 (48,49). 339 We further identified ERAD-C as the most likely pathway responsible for FIT2 340 degradation with increased MARCH6 and FIT2 protein association in the presence of 341 palmitate and reduced FIT2 degradation with MARCH6 knockdown. Mutating 342 predicted S-acylated cysteine residues abrogated FIT2 palmitoylation and degradation. 343 Further molecular work correlating degree of FIT2 palmitoylation with its degradation 344 in both physiology and pathology may help improve our understanding on how FIT2 345 function is fine-tuned at the protein level. Mouse pancreatic islets were isolated by perfusing the pancreas through the common 375 bile duct with collagenase as previously described (50) Cells and isolated islets were lysed in RIPA buffer supplemented with protease 389 inhibitor cocktail. Proteins were separated by SDS-PAGE and transferred onto 390 nitrocellulose membranes. Blocking was performed at room temperature for 1 h in Tris-391 buffered saline (TBS) with 5% non-fat milk, followed by incubation with the different 392 primary antibodies (described above) in blocking buffer for either 1 h at room 393 temperature or overnight at 4 °C. After several washes with TBS containing 0.5% 394 Tween 20 (TBST), the membranes were incubated with secondary antibodies of anti-395 mouse/rabbit IgG/HRP (as appropriate) in TBS with 1% non-fat milk. Following several 396 washes, the protein bands were visualized using enhanced chemiluminescence (Cell 397 Signaling Technology) and quantified using ImageJ. 398

RNA extraction and quantitative RT-PCR 400
Total RNA was prepared from tissues or cells using the NucleoSpin RNA II kit 401 (Macherey-Nagel), or prepared from the isolated islets using the NucleoSpin RNA XS 402 kit (Macherey-Nagel).

RNA-seq library preparation and data processing 409
Pancreatic islets were isolated from 12-week old wild type (WT) and β-cell specific 410 FIT2 knockout (β-FIT2KO, KO) mice in quintuplicate. Isolated pancreatic islets then 411 cultured in complete CMRL medium overnight for recovery. Total RNA was harvested 412 using RNeasy Plus Mini Kit (Qiagen) followed by RNA-seq library construction. Genes 413 were considered to be significantly differentially expressed when false discovery rate 414 (FDR) ≤ 0.05, with FPKM ≥ 1 in one sample group retained for subsequent analysis. 415 GO clustering enrichment analysis was carried out on differential genes using the 416 Functional Annotation tool in DAVID version 6.8 (51, 52)  Cryosections of 10 µm thickness from fixed and cryopreserved pancreas were used 425 for immunofluorescence analysis. Sections were rinsed with TBS, permeabilized, and 426 blocked with 10% normal goat serum plus 0.2% Triton X-100 in TBS for 1 h at room 427 temperature and then incubated overnight with primary antibodies at 4°C in a 428 humidified atmosphere. After gentle washing with TBS and incubating with 429 fluorescence secondary antibodies for 1 h at room temperature, sections were 430 mounted with Vecta Mount solution (Vector Labs) and multiple Z-stack images were 431 obtained using confocal imaging (Leica) and subsequently quantified using Image-J). 432 Briefly, the average fluorescence intensity of the staining, within insulin-positive ROI, 433 was quantified, corrected by subtraction of average background intensity, and 434 normalized to its control. For LD staining, cryosections were stained with BODIPY 435 493/503 (0.01mg/ml) and DAPI (5 μg/ml) for 15 min at RT. For propidium iodide (PI) 436 staining, cells were incubated with 10 μg/ml PI and 5 μg/ml DAPI in medium for 1h at 437 37 °C followed by RT fixation. Images were captured using a SP8 confocal microscope 438 (Leica). BODIPY-positive puncta per cell and percentage of PI-positive cells were 439 quantified using Image-J Prism 7 (GraphPad). 440 441

Total islet and pancreatic insulin content 442
To determine islet insulin content, 10 isolated islets were washed twice with ice-cold 443 D-PBS and then lysed with RIPA buffer. Insulin and protein content of the lysate were 444 measured using Mouse Insulin ELISA (Mercodia) and BCA assay (ThermoFisher 445 Scientific), respectively. Total islet insulin content was normalized to total protein 446 content. To determine pancreatic insulin content, half of the whole pancreas was 447 collected and placed into 5 ml Acid-Ethanol solution (1.5% HCl in 70% EtOH) overnight 448 at −20°C. Tissue was then homogenized and incubated in the same solution overnight 449 at −20°C. Supernatant was collected by centrifugation at 3000 x g for 10 min at 4°C, 450 followed by neutralization with 1 M Tris pH 7.5 at 1:1 (vol/vol  was split into two tubes: "experimental" and "negative control". In the experimental 528 tube, proteins were treated with thioester cleavage reagent, which cleaves off acyl 529 groups from the protein, resulting in the exposure of a free thiol group. Treated 530 samples were then subjected to CAPTUREome™ resin for 2.5 h at room temperature, 531 where proteins with free thiols were captured by the resin. 50 µl of flowthroughs were 532 collected as the cleaved unbound fraction (cUF). S-acylated proteins, which were 533 captured by the CAPTUREome™ resin, were eluted by incubating the resin in 50 µl of 534 2x laemmli buffer at 60°C for 10 min (cleaved Bound Fraction (cBF)). In the negative 535 control tube, proteins were treated with acyl-preservation reagent, which preserves 536 acyl groups on the protein. Treated samples were then subjected to CAPTUREome™ 537 resin for 2.5 h at room temperature. 50 µl of flowthroughs were collected as preserved 538 unbound fraction (pUF). Proteins were eluted from the CAPTUREome™ resin by 539 incubation of resin with 50 µl of 2x laemmli buffer at 60°C for 10 min (preserved Bound 540 Fraction (pBF)). 541 542

Statistical analysis 543
Data tested for normality (Shapiro-Wilk test) are expressed as mean ± SEM. 544 Parametric analysis (using Student's t-test) was used to determine statistical 545 difference and P values < 0.05 were considered as statistically significant (Prism 7, 546 GraphPad). 547 548 DATA AVAILABILITY 549 550 RNA sequencing data reported in this paper is available at NCBI GEO Accession no.: 551 GSE133939. All data mentioned in this paper will be placed on a data repository. 552 Authors declare no primary datasets and computer codes linked to this study.